Nozzle maintenance support device and nozzle maintenance support program

Abstract

To minimize the frequency of maintenance while enabling maintenance of a nozzle to be performed before an abnormality occurs in the nozzle.SOLUTION: In a nozzle maintenance support device (component mounting device 3) for component suction provided on a component mounting device 3 that mounts a component E on a substrate P, a control unit executes a first estimation process to estimate the degree and type of abnormality of the nozzle on the basis of information on the nozzle, a determination process to determine whether to predict the remaining life of the nozzle for the type of abnormality estimated in the first estimation process by comparing the degree of abnormality estimated in the first estimation process with a second threshold value, and a prediction process to predict the remaining life of the nozzle until a suction error or mounting failure caused by the nozzle occurs, on the basis of a remaining life evaluation value, for the type of abnormality determined in the determination process to predict the remaining life.SELECTED DRAWING: Figure 2

Description

[Technical field]

[0001] The technology disclosed in this specification relates to a nozzle maintenance support device and a nozzle maintenance support program. [Background technology]

[0002] A nozzle for suctioning components that is provided in a component mounting device may develop some abnormality during use, and therefore, a method for detecting nozzle abnormalities is performed (see, for example, Patent Document 1). Specifically, the surface mount machine described in Patent Document 1 includes a nozzle shaft having a supply path that supplies air to the inside, a filter attached to the supply path, and a suction nozzle attached to the tip of the nozzle shaft, and when there is an abnormality in the negative pressure in the mounting head, the combination of the suction nozzle and nozzle shaft is changed and the pressure or air flow rate of the supply path is remeasured, and based on the measurement results obtained, it is determined whether the cause of the negative pressure abnormality is a clogged suction nozzle or a clogged filter. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] JP 2020-202306 A Summary of the Invention [Problem to be solved by the invention]

[0004] However, the surface mounter described in Patent Document 1 determines the detailed cause (or location) of an abnormality that has already occurred. Therefore, the surface mounter described in Patent Document 1 has room for improvement in terms of minimizing the frequency of nozzle maintenance. This specification discloses a technique that enables nozzle maintenance to be performed before an abnormality occurs in the nozzle, while minimizing the frequency of that maintenance. [Means for solving the problem]

[0005] The nozzle maintenance support device disclosed herein is a maintenance support device for a component suction nozzle provided in a component mounting device that mounts components on a board, and includes an acquisition unit that acquires information about the nozzle, and a processing unit, wherein the processing unit executes: a first estimation process that estimates the degree and type of abnormality of the nozzle based on the information acquired by the acquisition unit; a judgment process that determines whether or not to predict the remaining life of the nozzle for the type of abnormality estimated in the first estimation process by comparing the degree of abnormality estimated in the first estimation process with a threshold value; and a prediction process that predicts the remaining life of the nozzle until a suction error or mounting failure caused by the nozzle occurs, for the type of abnormality determined to have a remaining life predicted in the judgment process, based on the remaining life evaluation value, when the degree of abnormality estimated in the first estimation process is defined as a remaining life evaluation value, or when an evaluation value estimated based on the information acquired by the acquisition unit and for predicting the remaining life of the nozzle is defined as a remaining life evaluation value. Effect of the Invention

[0006] According to the above configuration, it is possible to perform nozzle maintenance before an abnormality occurs in the nozzle, while minimizing the frequency of the maintenance. [Brief description of the drawings]

[0007] [Figure 1] Block diagram of a component mounting system according to embodiment 1. [Diagram 2] Top view of component mounting equipment [Diagram 3] Schematic diagram of the head unit seen from the front [Figure 4] A block diagram showing the electrical configuration of a component mounting device. [Diagram 5] A block diagram showing the electrical configuration of a management device. [Figure 6] Schematic diagram to explain types of nozzle abnormalities [Figure 7]Flowchart of a process for estimating the degree of abnormality [Figure 8] A graph showing the relationship between the number of shots and the remaining life evaluation value, and a graph showing the relationship between the number of shots and the occurrence rate of mounting defects. [Figure 9] Nozzle health check flow chart [Figure 10] Flowchart of prediction and notification processing [Figure 11] A sequence chart for storing sensor information in a storage unit of a management device according to a second embodiment. [Figure 12] Schematic diagram of a side-view camera according to a third embodiment. [Figure 13] Nozzle health check flow chart [Figure 14A] Flowchart of the process for re-estimating the degree of abnormality in the case of material damage / wear-surface-chip [Figure 14B] Flowchart of a process for determining whether an abnormality occurs in the case of dirt-surface-solder [Figure 14C] Flowchart of the process of re-estimating the degree of abnormality in the case of dirt - hole (clogged hole) - "dust, lint" [Figure 15] 13 is a graph showing the relationship between the number of shots and the remaining life evaluation value in a re-estimation example according to the fourth embodiment. [Figure 16] Graph showing the relationship between the number of shots and remaining life evaluation value in correction example 1 [Figure 17] Graph showing the relationship between the number of shots and remaining life evaluation value in correction example 2 [Figure 18] Schematic diagram of a correction screen according to embodiment 6 [Figure 19] Flowchart of classifier re-learning process [Figure 20] Schematic diagram of the inquiry screen [Figure 21] 13 is a graph showing the relationship between the number of shots and the remaining life evaluation value according to the seventh embodiment, and a graph showing the relationship between the number of shots and the incidence rate of mounting defects. [Figure 22] Learning process flowchart [Figure 23] Block diagram of a component mounting system according to an eighth embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0008] [Overview of the embodiment of the present disclosure] (1) A nozzle maintenance support device according to an embodiment is a maintenance support device for a component suction nozzle provided in a component mounting device that mounts components on a board, and includes an acquisition unit that acquires information about the nozzle, and a processing unit, wherein the processing unit executes: a first estimation process that estimates the degree and type of abnormality of the nozzle based on the information acquired by the acquisition unit; a judgment process that determines whether or not to predict the remaining life of the nozzle for the type of abnormality estimated in the first estimation process by comparing the degree of abnormality estimated in the first estimation process with a threshold value; and a prediction process that predicts the remaining life of the nozzle until a suction error or mounting failure caused by the nozzle occurs, for the type of abnormality determined to have a remaining life predicted in the judgment process, based on the remaining life evaluation value, when the degree of abnormality estimated in the first estimation process is defined as a remaining life evaluation value, or when an evaluation value estimated based on the information acquired by the acquisition unit and for predicting the remaining life of the nozzle is defined as a remaining life evaluation value.

[0009] In the case of the surface mounter described in Patent Document 1 mentioned above, where a nozzle abnormality cannot be detected until after a pickup error or mounting failure has occurred, maintenance is often performed on a time-managed basis, such as once every three days or once a week, to prevent nozzle abnormalities from occurring. However, with time-managed maintenance, there is a possibility that maintenance will be performed even when the degree of the nozzle abnormality is minor and therefore maintenance is not yet necessary. This could result in a decrease in operator work efficiency due to unnecessary maintenance.

[0010] According to the nozzle maintenance support device described in (1) above, the remaining lifespan of a nozzle until a pickup error or mounting failure due to the nozzle occurs is predicted, so it is possible to, for example, notify the operator of the predicted remaining lifespan, or, without notifying the remaining lifespan itself, to notify the operator of a guide such as "The nozzle will soon reach the end of its lifespan. Please perform maintenance." This enables the operator to perform nozzle maintenance before an abnormality occurs in the nozzle. By providing the above-mentioned notification, even if a long time has passed since the last maintenance, the operator will not carry out maintenance unless notified of the remaining life or maintenance guidance, which can reduce unnecessary nozzle maintenance compared to when maintenance is performed based on time management. This makes it possible to minimize the frequency of maintenance. Therefore, according to the nozzle maintenance support device described in (1) above, it is possible to perform nozzle maintenance before an abnormality occurs in the nozzle, while minimizing the frequency of the maintenance.

[0011] There are various types of nozzle abnormalities. The remaining life from when an abnormality begins to occur until the abnormality actually occurs differs depending on the type of abnormality. For this reason, if there is an abnormality where the nozzle hole becomes clogged and causes a suction error, and an abnormality where the bottom surface of the nozzle becomes dirty and causes a suction error, for example, it is preferable to predict the remaining life according to each abnormality. Previously, attempts had been made to detect signs of nozzle abnormalities, but these attempts did not take into account the type of abnormality and instead attempted to detect signs when a variety of types of abnormalities were mixed together, resulting in low detection accuracy. According to the nozzle maintenance support device described above in (1), the remaining life is predicted for each type of nozzle abnormality, making it possible to predict the remaining life with greater accuracy than when predicting the remaining life in a state where various types of abnormalities are mixed.

[0012] (2) In the nozzle maintenance support device described in (1) above, the processing unit may execute a first notification process of notifying an operator of the remaining life predicted in the prediction process via a notification unit.

[0013] According to the nozzle maintenance support device described in (2) above, the remaining life of the nozzle is notified to the operator, so that the operator can determine whether or not the nozzle should be maintained. Therefore, unnecessary nozzle maintenance can be reduced compared to when maintenance is performed based on time management.

[0014] (3) In the nozzle maintenance support device described in (2) above, in the first notification process, the processing unit may change a notification method depending on the remaining life predicted in the prediction process.

[0015] According to the nozzle maintenance support device described in (3) above, for example, when the remaining life is short (for example, when the remaining life is equal to or less than a set value), the remaining life of the nozzle is notified to the operator in a manner that reliably conveys the remaining life to the operator, thereby increasing the likelihood that the operator will immediately perform nozzle maintenance. On the other hand, when the remaining life is long, the notification is made in a manner that is less likely to interfere with the operator's work, thereby preventing the notification of the remaining life from interfering with the operator's work.

[0016] (4) In the nozzle maintenance support device described in any one of (1) to (3) above, the processing unit may execute a second notification process to notify an operator via a notification unit of a maintenance method corresponding to the type of abnormality for which the remaining life has been predicted in the prediction process.

[0017] As the components to be mounted become finer, more frequent maintenance of nozzles is required. In addition, as nozzles become finer, more precise maintenance work is required, which tends to extend the maintenance time. For this reason, automatic cleaning machines that automatically clean nozzles are being developed, but automatic cleaning machines cannot clean difficult dirt such as solder that has adhered to the nozzle surface.

[0018] According to the nozzle maintenance support device described in (4) above, the maintenance method according to the type of abnormality is notified, so for example, in the case of a type of abnormality that can be maintained with an automatic cleaning machine, maintenance using the automatic cleaning machine is notified as the maintenance method, making it possible to make effective use of the automatic cleaning machine and reducing work requiring manpower as much as possible. This makes it possible to perform maintenance with the minimum amount of effort required.

[0019] (5) In the nozzle maintenance support device described in any one of (1) to (3) above, the processing unit may execute a third notification process to notify an operator via the notification unit that an abnormality has occurred in the nozzle that may affect pickup errors or mounting defects when the remaining life predicted by the prediction process is equal to or less than a set value.

[0020] When the remaining life is equal to or less than a set value, it is preferable to perform maintenance immediately. According to the nozzle maintenance support device transferred as described above in (5), when the remaining life is equal to or less than a set value, the operator is notified, thereby increasing the possibility that the operator will perform nozzle maintenance immediately.

[0021] (6) A nozzle maintenance support device as described in any one of (1) to (3) above, wherein the nozzle maintenance support device is a computer communicatively connected to the component mounting device, and the acquisition unit may acquire the information by receiving the information from the component mounting device.

[0022] When realizing a nozzle maintenance support device, a component mounting device can function as the nozzle maintenance support device, or a computer that is communicatively connected to the component mounting device can function as the nozzle maintenance support device.

[0023] In order to accurately predict the remaining life of a nozzle, the processing unit must handle a large amount of data, and thus the processing unit must have high processing power. Therefore, when a component mounting device is used as a nozzle maintenance support device, each component mounting device must be equipped with a processing unit with high processing power, which increases the cost of the component mounting device. According to the nozzle maintenance support device described in (6) above, a computer that is communicatively connected to the component mounting device functions as the nozzle maintenance support device, and the computer receives information from the component mounting device, making it possible to predict the remaining lifespan of each component mounting device without equipping each device with a processing unit with high processing power.

[0024] When a component mounting device with a low processing capacity of a processing section is made to function as a nozzle maintenance support device, the processing section is loaded with a high load due to the processing section also functioning as a nozzle maintenance support device. This may result in a decrease in the production efficiency of the component mounting device due to a waiting time until the production of the board starts. According to the nozzle maintenance support device described in (6) above, by making a computer separate from the component mounting device function as the nozzle maintenance support device, it is possible to suppress a decrease in the production efficiency of the component mounting device.

[0025] (7) In the nozzle maintenance support device described in (6) above, the processing unit executes an acquisition process to acquire additional information for estimating the evaluation value via the acquisition unit, and a second estimation process to estimate the evaluation value based on the additional information acquired by the acquisition unit, and in the prediction process, may predict remaining life based on the evaluation value estimated in the second estimation process.

[0026] According to the nozzle maintenance support device described in (7) above, it is usually possible to determine with a small amount of processing whether to predict the remaining life by estimating the degree of abnormality with a small amount of information. And, according to the nozzle maintenance support device, for the type of abnormality for which it is determined that the remaining life should be predicted, additional information is obtained from the component mounting device and an evaluation value is estimated, thereby making it possible to predict the remaining life with high accuracy.

[0027] (8) A nozzle maintenance support device as described in any one of (1) to (3) above, comprising: a first storage process for storing the estimated remaining life evaluation value in a memory unit; an estimation process for estimating a current remaining life evaluation value based on the past remaining life evaluation values ​​stored in the memory unit; and a re-estimation process for re-acquiring information about the nozzle and re-estimating the current remaining life evaluation value if a difference between the currently estimated remaining life evaluation value and the current remaining life evaluation value estimated in the estimation process is equal to or greater than a reference value, and wherein in the prediction process, the remaining life of the nozzle is predicted based on the remaining life evaluation value re-estimated in the re-estimation process.

[0028] According to the nozzle maintenance support device described in (8) above, by eliminating sudden outliers (failure to estimate the remaining life evaluation value), the remaining life evaluation value can be estimated more accurately (and thus the remaining life can be predicted with greater precision).

[0029] (9) A nozzle maintenance support device as described in any one of (1) to (3) above, wherein the processing unit executes a first storage process for storing the remaining life evaluation value in a memory unit, and a correction process for correcting the remaining life evaluation value currently estimated based on the past remaining life evaluation value stored in the memory unit, and in the prediction process, predicts the remaining life of the nozzle based on the remaining life evaluation value corrected by the correction process.

[0030] According to the nozzle maintenance support device described in (9) above, the currently estimated remaining life evaluation value is corrected based on the previously estimated remaining life evaluation value, thereby improving the accuracy of remaining life prediction.

[0031] (10) In the nozzle maintenance support device described in any one of (1) to (3) above, the processing unit may estimate the degree and type of abnormality using a classifier created by unsupervised learning.

[0032] There are various types of nozzles, differing in the size, shape, number of nozzle holes, etc. When creating a classifier through machine learning using training data, if one were to try to learn the degree of anomaly for each type of anomaly for all nozzle types, it would be necessary to prepare training data that covers all types of anomaly and all degrees of anomaly for all nozzle types, which is difficult to achieve. According to the nozzle maintenance support device described in (10) above, since the classifier is created by unsupervised learning (including deep learning), the classifier can be created with little effort.

[0033] (11) The nozzle maintenance support device described in (10) above may include an operation unit, and the processing unit may execute a correction acceptance process that accepts from an operator via the operation unit a correction of at least one of the degree of abnormality and the type of abnormality estimated by the classifier, and a re-learning process that re-learns the classifier based on the correction result accepted in the correction acceptance process.

[0034] According to the nozzle maintenance support device described in (11) above, if the estimation result of the classifier is incorrect, the operator corrects the estimation result and the classifier is re-learned using the corrected result, thereby improving the reliability of the estimation result.

[0035] (12) In the nozzle maintenance support device described in any one of (1) to (3) above, the processing unit may execute a learning process to learn the remaining life evaluation value for determining that the remaining life is 0, based on the remaining life evaluation value at the time when maintenance is performed on the nozzle before a suction error or mounting defect caused by the nozzle occurs.

[0036] According to the nozzle maintenance support device described in (12) above, the remaining life evaluation value at which the remaining life is determined to be zero is learned, thereby improving the accuracy of predicting the remaining life.

[0037] (13) In the nozzle maintenance support device described in any one of (1) to (3) above, the processing unit may execute a second storage process that associates nozzle identification information of the nozzle with the remaining life evaluation value and stores them in a data storage server.

[0038] A nozzle equipped in a component mounting device may be moved to another component mounting device for use. For example, if a remaining life evaluation value for a nozzle equipped in a component mounting device is estimated by a maintenance support device for that nozzle, and a remaining life evaluation value for a nozzle equipped in another component mounting device is estimated by a maintenance support device for another nozzle, the estimated remaining life evaluation values ​​are stored in a distributed manner in the maintenance support devices for the multiple nozzles, making it difficult to use all remaining life evaluation values ​​previously estimated for a certain nozzle in any processing.

[0039] According to the nozzle maintenance support device described in (13) above, the estimated remaining life evaluation value is stored in a data storage server, so that regardless of which nozzle maintenance support device estimates the remaining life evaluation value, it becomes easy to use all remaining life evaluation values ​​previously estimated for a certain nozzle for some kind of processing.

[0040] (14) A nozzle maintenance support program according to an embodiment is a maintenance support program for a component suction nozzle provided in a component mounting device that mounts components on a board, and causes a computer to execute the following: a first estimation process that estimates the degree and type of abnormality of the nozzle based on information about the nozzle; a judgment process that determines whether to predict the remaining life of the nozzle for the type of abnormality estimated in the first estimation process by comparing the degree of abnormality estimated in the first estimation process with a threshold value; and a prediction process that, when the degree of abnormality estimated in the first estimation process is defined as a remaining life evaluation value, or when an evaluation value estimated based on information about the nozzle and for predicting the remaining life of the nozzle is defined as a remaining life evaluation value, predicts the remaining life of the nozzle until a suction error or mounting failure caused by the nozzle occurs for the type of abnormality determined to have a remaining life predicted in the judgment process based on the remaining life evaluation value.

[0041] According to the maintenance support program described in (14) above, it is possible to perform nozzle maintenance before an abnormality occurs in the nozzle, while minimizing the frequency of the maintenance.

[0042] [Details of the embodiment of the present disclosure] The present disclosure will be described in detail below with reference to the exemplary embodiments. The present disclosure is not limited to these examples, but is defined by the claims, and is intended to include all modifications within the meaning and scope of the claims. The embodiments of the present disclosure can be realized in various forms, such as an apparatus, a method, a computer program for implementing the functions of these apparatuses or methods, and a recording medium on which the computer program is recorded.

[0043] <Embodiment 1> In the following description, the left-right direction in Fig. 2 is referred to as the X direction, the front-rear direction is referred to as the Y direction, the up-down direction in Fig. 3 is referred to as the Z direction, and the right side in Fig. 2 is referred to as the upstream side and the left side is referred to as the downstream side. In the following description, the reference numerals in the drawings may be omitted for the same components, with some exceptions.

[0044] (1) Component Mounting System A component mounting system 1 according to a first embodiment will be described with reference to Fig. 1. The component mounting system 1 is a system that mounts a component E (see Fig. 2) such as an electronic component on a substrate P (see Fig. 2). The component mounting system 1 includes a management device 2 (an example of a computer), a component mounting device 3 (an example of a nozzle maintenance support device), and an automatic cleaning machine 4. These are connected to each other via a communication network 5 so as to be able to communicate with each other.

[0045] (2) Configuration of component mounting equipment As shown in FIG. 2, the component mounting apparatus 3 includes a stand 31, a transport conveyor 32, four tape component supply devices 33, a head unit 34, a head moving unit 35, two component imaging cameras 36 (an example of an acquisition unit), a board imaging camera 37, a nozzle station 38, a control unit 60 (see FIG. 4) and an operation panel 61 (see FIG. 4). When the component mounting apparatus 3 functions as a nozzle maintenance support device, the control unit 60 of the component mounting apparatus 3 is an example of a processing unit. The operation panel 61 is an example of a notification unit and an operation unit.

[0046] 2, an area A indicated by a dashed double-dashed line is a work position (hereinafter referred to as work position A) where the board P is fixed when mounting components E on the board P. The transport conveyor 32 is a device that transports the board P carried in from the upstream side to the work position A, and carries the board P on which the components E are mounted at the work position A out to the downstream side. The transport conveyor 32 is equipped with a pair of conveyor belts 39 (a front conveyor belt 39A and a rear conveyor belt 39B) that circulate in the X direction, a plurality of rollers around which the conveyor belt 39 is stretched, a conveyor drive motor 64 (see FIG. 4) that drives and rotates the rollers, and the like.

[0047] The tape component supply devices 33 are arranged in a row in the X direction on both sides in the Y direction of the component mounting device 3, for a total of four locations. A plurality of tape feeders 33A are attached to the tape component supply device 33 and aligned horizontally in the X direction. Each tape feeder 33A is equipped with a reel around which a component tape holding components E is wound, and an electric feeding device that unwinds the component tape from the reel, and supplies components E one by one.

[0048] The head unit 34 is equipped with a plurality of mounting heads 41 that pick up and release the components E. The head unit 34 is of a so-called in-line type, and the plurality of mounting heads 41 are arranged side by side in the X direction. The head unit 34 will be described later. Here, an in-line type head unit 34 will be described as an example of the head unit 34, but the head unit 34 may also be a so-called rotary head in which the plurality of mounting heads 41 are arranged on a circumference.

[0049] Head moving section 35 is a mechanism for moving head unit 34 in the X and Y directions within a predetermined movable range. Head moving section 35 includes a beam 42 supporting head unit 34 so as to be reciprocatable in the X direction, a pair of Y-axis guide rails 43 supporting beam 42 so as to be reciprocatable in the Y direction, an X-axis servo motor 44 (see FIG. 4) for moving head unit 34 back and forth in the X direction, and a Y-axis servo motor 45 (see FIG. 4) for moving beam 42 back and forth in the Y direction.

[0050] The two component imaging cameras 36 are provided between the two tape component supply devices 33 aligned in the X direction. The component imaging cameras 36 image, from below, the components E that are sucked onto the mounting head 41. The component imaging cameras 36 are also used to image the underside of the nozzle 41C (see FIG. 3). The board imaging camera 37 is a camera that captures from above an image of a fiducial mark marked on the top surface of the board P. The board imaging camera 37 is attached to the head unit 34, and moves in the X and Y directions together with the head unit 34. As will be described in more detail later, the board imaging camera 37 is also used to recognize the nozzle 41C stored in the nozzle station 38.

[0051] Nozzle station 38 is equipped with a nozzle plate having multiple storage holes in which nozzles 41C (see FIG. 3) for suctioning components are stored, a shutter arranged on the nozzle plate, a shutter movement mechanism for moving the shutter in a direction parallel to the nozzle plate, etc. When the shutter is open, nozzle 41C can be removed from the nozzle plate, and when the shutter is closed, it cannot be removed.

[0052] The head unit 34 will be described with reference to Fig. 3. The head unit 34 is provided with a plurality of mounting heads 41, a Z-axis servo motor 62 (see Fig. 4) which raises and lowers these mounting heads 41 individually, and an R-axis servo motor 63 (see Fig. 4) which rotates these mounting heads 41 all at once around their axis.

[0053] The mounting head 41 includes a nozzle shaft 41A, a spring 41B, and a nozzle 41C detachably attached to the lower end of the nozzle shaft 41A. The nozzle shaft 41A is cylindrical and supported by the head unit 34 so as to be movable up and down. The nozzle shaft 41A is connected to an external air supply device via an air supply path, and positive and negative pressures are supplied to the inside of the nozzle shaft 41A by the external air supply device. A pressure sensor (an example of an acquisition unit) that measures the positive and negative pressures is provided in the air supply path.

[0054] Nozzle shaft 41A is inserted through coil spring 41B. When nozzle shaft 41A is pressed down by Z-axis servo motor 62, spring 41B expands. When nozzle shaft 41A is no longer pressed by Z-axis servo motor 62, the force of expanded spring 41B returning to its original state causes nozzle shaft 41A to rise. Grease is applied between nozzle shaft 41A and spring 41B to reduce friction.

[0055] Negative pressure and positive pressure are supplied to the nozzle 41C from an air supply device via the nozzle shaft 41A. The nozzle 41C sucks the component E when negative pressure is supplied thereto, and releases the component E when positive pressure is supplied thereto. Nozzle identification information for uniquely identifying nozzle 41C is written on the top surface of nozzle 41C. Component mounting device 3 captures an image of nozzle 41C stored in nozzle station 38 from above with board imaging camera 37, and detects nozzle 41C stored in nozzle station 38 by recognizing the nozzle identification information on the captured image.

[0056] (3) Electrical configuration of the component mounting device 4, the electrical configuration of the component mounting device 3 will be described. The component mounting device 3 includes a control unit 60 and an operation panel 61. The control unit 60 includes an arithmetic processing unit 60A, a motor control unit 60B, a storage unit 60C, an image processing unit 60D, an external input / output unit 60E, a feeder communication unit 60F, and a communication unit 60G.

[0057] The arithmetic processing unit 60A includes a CPU, a RAM, and the like, and controls each unit of the component mounting apparatus 3 by executing a control program stored in the storage unit 60C. The motor control unit 60B controls the rotation of each motor such as the X-axis servo motor 44 and the Y-axis servo motor 45 under the control of the arithmetic processing unit 60A.

[0058] The storage unit 60C stores various programs and data executed by the calculation processing unit 60A. The various programs include a nozzle maintenance support program. The data stored in the storage unit 60C includes component mounting data in which the target mounting position, target mounting angle, mounting order, etc. of each component E are set, and various data referenced by the nozzle maintenance support program. The image processing section 60D is configured to receive image signals output from the component imaging camera 36 and the board imaging camera 37.

[0059] The external input / output unit 60E is a so-called interface, and is configured to receive detection signals output from various sensors 65 provided in the main body of the component mounting apparatus 3. The external input / output unit 60E is also configured to control the operation of various actuators 66 (such as an air supply device) based on control signals output from the arithmetic processing unit 60A.

[0060] Feeder communication unit 60F is connected to tape feeder 33A and controls tape feeder 33A in an integrated manner. The communication unit 60G is a communication circuit for connecting the component mounting apparatus 3 to the communication network 5. The operation panel 61 includes a display device such as a liquid crystal display, and an input device. The input device is configured with a touch panel arranged on the front side of the display device, various operation buttons, and the like.

[0061] (4) Management device The electrical configuration of the management device 2 will be described with reference to Fig. 5. The management device 2 is a so-called personal computer, and includes a processing unit 70, a memory unit 71, a display unit 72, an operation unit 73, and a communication unit 74. The processing unit 70 has a CPU 70A and a RAM 70B. The display unit 72 is composed of a display device such as a liquid crystal display, a drive circuit for driving the display device, and the like. The operation unit 73 includes a keyboard, a mouse, and a touch panel. The communication unit 74 is a communication interface for connecting the management device 2 to the communication network 5. The memory unit 71 stores various programs and data executed by the CPU 70A.

[0062] (5) Nozzle health check The nozzle health check is a process for checking the state of the nozzle 41C. In the first embodiment, the nozzle health check is performed periodically. For example, the nozzle health check is performed every day before the component mounting apparatus 3 starts operating. The timing for performing the nozzle health check can be determined appropriately. For example, the component mounting apparatus 3 may perform the nozzle health check when instructed by an operator.

[0063] As will be described in detail later, there are various types of abnormalities in the nozzle 41C. In the nozzle health check, the control unit 60 acquires information (hereinafter referred to as sensor information) about all of the nozzles 41C stored in the nozzle station 38 using the acquisition unit (the component imaging camera 36 and the pressure sensor). The sensor information includes an image of the bottom surface of the nozzle 41C captured from below by the component imaging camera 36, ​​pressure measured by a pressure sensor that measures the positive pressure and negative pressure supplied to the nozzle 41C, etc.

[0064] The control unit 60 estimates the degree and type of abnormality of the nozzle 41C based on the acquired sensor information (first estimation process), and for the estimated type of abnormality, compares the estimated degree of abnormality with a second threshold value (an example of a threshold value) described later to determine whether or not to predict the remaining life of the nozzle 41C (an example of a determination process). Here, the degree of abnormality refers to how far it deviates from a normal state. Then, when the estimated degree of abnormality is defined as the remaining life evaluation value, or when the evaluation value estimated based on the information acquired by the acquisition unit and used to predict the remaining life of the nozzle 41C is defined as the remaining life evaluation value, the control unit 60 predicts the remaining life of the nozzle 41C until an adsorption error or mounting failure caused by the nozzle 41C occurs, for the type of abnormality for which it has determined that the remaining life is to be predicted, based on the remaining life evaluation value (an example of a prediction process).

[0065] Here, a mounting failure refers to a situation where the nozzle 41C was able to pick up the component E, but the picked up component E could not be properly mounted on the board P. Examples of mounting failures include a misalignment of the position or angle of the component E mounted on the board P, or the component E picked up by the nozzle 41C being taken away without being mounted on the board P (so-called component take-away), etc. A suction error refers to an error in which the nozzle 41C fails to pick up the component E, or an error in which the component E picked up by the nozzle 41C falls midway and is not mounted on the board P. If a suction error occurs, the component mounting device 3 retries mounting the component E a predetermined number of times. If the component E can be mounted after the retry, it is not considered a mounting failure. For this reason, in the embodiment, a distinction is made between a suction error and a mounting failure. If the component E cannot be mounted even after the predetermined number of retries, it is considered a mounting failure.

[0066] The control unit 60 according to the first embodiment also executes a notification process for notifying an operator of information based on the remaining life predicted in the prediction process. Specifically, the control unit 60 executes the following first to third notification processes.

[0067] The first notification process is a process of notifying an operator of the remaining life predicted in the prediction process via the operation panel 61. Although details will be described later, in the first notification process, the control unit 60 changes the notification method depending on the remaining life predicted in the prediction process. The second notification process is a process of notifying the operator via the operation panel 61 of a maintenance method corresponding to the type of abnormality for which the remaining life has been predicted in the prediction process. The third notification process is a process for notifying the operator via the operation panel 61 that an abnormality has occurred in the nozzle 41C that may affect pickup errors or mounting defects when the remaining life predicted by the prediction process is equal to or less than a set value.

[0068] Here, methods for estimating the degree and type of abnormality of the nozzle 41C include learning-based estimation using machine learning and non-learning-based estimation not using machine learning. In the first embodiment, a non-learning-based estimation not using machine learning will be described as an example.

[0069] (5-1) Types of Abnormality The types of abnormalities in the nozzle 41C will be described with reference to Fig. 6. The types of abnormalities are classified according to three attributes: nature, location, and details.

[0070] The nature is classified as dirt or damage / wear. If the nature is dirt, the part is classified as surface or hole (hole clogging). Here, surface refers to the bottom surface of nozzle 41C. When the nature is dirt and the location is surface, the details are classified as solder, electrode, or "dust, lint." Solder is an abnormality in which solder printed on the printed circuit board P adheres to the underside of the nozzle 41C. Electrode is an abnormality in which an electrode on the component E has been scraped off and adheres to the underside of the nozzle 41C. "Dust, lint" is an abnormality in which dust or lint has adhered inside the component mounting device 3. When the nature is dirt and the location is a hole (hole clogged), the details are classified as "dust, lint" or embossed pieces. "Dust, lint" is an abnormality where dust or lint has gotten into the nozzle hole, causing it to become clogged. "Embossed pieces" is an abnormality where a piece of cover tape (embossed pieces) from the component tape has gotten into the nozzle hole, causing it to become clogged.

[0071] If the property is damage / wear, the part is classified as surface or other. If the nature is damage / wear and the location is surface, the details are classified as cracks or chips. A crack is an abnormality where the tip surface of the nozzle 41C is cracked. A chip is an abnormality where the tip of the nozzle 41C is chipped. If the nature is physical damage / wear and the location is other than that, the details are classified as breakage or spring / grease deterioration. A breakage is an abnormality in which the nozzle 41C is broken. A spring / grease deterioration is an abnormality in which the nozzle shaft 41A (in other words, the nozzle 41C) becomes difficult to move up and down due to deterioration of the spring 41B or grease provided on the nozzle shaft 41A.

[0072] The occurrence type indicates whether the type of abnormality is sudden or cumulative. A sudden type abnormality occurs suddenly without warning. From past experience, sudden abnormalities tend to cause pickup errors or mounting failures immediately even if the degree of abnormality is small. For this reason, the control unit 60 predicts that the remaining life of nozzle 41C is 0 (zero) when the type of abnormality is sudden. In other words, the control unit 60 determines that nozzle 41C has reached the end of its life.

[0073] An accumulation type abnormality is an abnormality that gradually accumulates. In the case of an accumulation type abnormality, even if it occurs, there is a tendency that a pickup error or mounting failure does not occur immediately. Therefore, when the type of abnormality is an accumulation type, the control unit 60 predicts the remaining life of the nozzle 41C until a pickup error or mounting failure starts to occur based on a remaining life evaluation value described later. The urgency level is the urgency of dealing with the abnormality. A sudden type has a high urgency level, while a cumulative type has a low urgency level.

[0074] The countermeasures are maintenance methods for each type of abnormality. The countermeasures for each type of abnormality are explained below. In the case of dirt-surface-solder, the solution to be taken is to "remove and clean by hand." In the case of dirt-surface-electrode, the notification "Remove and clean by hand (or replace)" is displayed. In the case of dirt-surface-"dust, lint", "automatic cleaner" is notified. Automatic cleaner 4 is a device that automatically cleans nozzle 41C. In the case of dirt-hole (clogged hole-"dust, lint" and dirt-hole (clogged hole-"dust, lint", "Air blow at maximum power, automatic cleaning machine" is notified.

[0075] In the cases of physical damage / wear-surface-crack, physical damage / wear-surface-chip, and physical damage / wear-other-break, "replacement" is reported. In the case of damage / wear-other-spring / grease, a notification to "replace / inject grease" will be displayed. The image is an example of the bottom surface of nozzle 41C captured by part imaging camera 36.

[0076] (5-2) Severity of abnormality The degree of abnormality may be estimated individually for each type of abnormality, or a single overall (overall) degree of abnormality for one nozzle 41C may be estimated. For convenience, the degree of abnormality for each type of abnormality is referred to as the individual degree of abnormality, and the overall degree of abnormality for one nozzle 41C is referred to as the common degree of abnormality.

[0077] (5-2-1) Degree of individual abnormality The scale of abnormality varies depending on the type of abnormality. Below, the scale of abnormality for accumulation type abnormality will be explained.

[0078] In the cases of soiling-surface-electrode and soiling-surface-"dust, lint", the degree of abnormality is expressed by the ratio (%) of the soiled area to the area of ​​the bottom surface of the nozzle 41C.

[0079] Dirt - Holes (clogged holes) - In the case of "dust, lint," the degree of abnormality is expressed by the difference between the negative pressure when the mounting head 41 is lowered and the nozzle 41C is pressed against the component E, and the negative pressure when the nozzle 41C is not pressed against the component E. Specifically, when the nozzle 41C is pressed against the component E, the negative pressure is high because air is not easily sucked into the nozzle 41C regardless of whether the nozzle 41C is clogged. When the nozzle 41C is not pressed against the component E, if the nozzle 41C is not clogged, sufficient air is sucked into the nozzle 41C, causing the negative pressure to decrease, increasing the difference from the negative pressure when the nozzle 41C is pressed against the component E. In contrast, when the nozzle 41C is clogged, sufficient air is not sucked into the nozzle 41C, causing the negative pressure to increase, decreasing the difference from the negative pressure when the nozzle 41C is pressed against the component E. Therefore, the degree of abnormality is represented by the difference between the negative pressure when the nozzle 41C is pressed against the component E and the negative pressure when it is not pressed against the component E.

[0080] In the case of physical damage / wear-surface-crack, the degree of abnormality is expressed, for example, by the length (mm) of the crack. The degree of abnormality may be expressed not by the length of the crack, but by the ratio (%) of the cracked area to the area of ​​the bottom surface of the nozzle 41C.

[0081] In the case of physical damage / wear-other-spring / grease deterioration, the degree of abnormality is indicated by the negative pressure at a predetermined time after the mounting head 41 starts to rise, when the mounting head 41 is lowered to press the nozzle 41C against the component E and then raised. Specifically, when the spring / grease is not deteriorated, the nozzle shaft 41A and the spring 41B slide smoothly and the mounting head 41 rises quickly, so that a sufficient gap is created between the bottom surface of the nozzle 41C and the top surface of the component E when a predetermined time has elapsed since the mounting head 41 started to rise. As a result, sufficient air is sucked into the inside of the nozzle 41C, and the negative pressure decreases. In contrast, when the spring / grease is deteriorated, the nozzle shaft 41A does not rise quickly, so that a sufficient gap is not created between the bottom surface of the nozzle 41C and the top surface of the component E when a predetermined time has elapsed. As a result, sufficient air is not sucked into the inside of the nozzle 41C, and the negative pressure increases. Therefore, the degree of abnormality is represented by the negative pressure when a predetermined time has elapsed since the mounting head 41 started to rise.

[0082] (5-2-2) Degree of common abnormalities As described above, the scale of the degree of abnormality differs depending on the type of abnormality. Therefore, even when estimating the degree of a common abnormality, the control unit 60 estimates the degree of an individual abnormality for each type of abnormality. Then, the control unit 60 estimates the degree of the common abnormality by adding the degrees of the individual abnormalities estimated for each type of abnormality with a weighting according to the type of abnormality, or by averaging the degrees with a weighting according to the type of abnormality.

[0083] (5-3) First estimation process A first estimation process for estimating the degree and type of abnormality of nozzle 41C will be described. Here, the process will be divided into a process for estimating the degree of abnormality and a process for estimating the type of abnormality. Here, the degree of each individual abnormality will be described as an example.

[0084] (5-3-1) Estimation of the degree of abnormality The flow of a process for estimating the degree of abnormality will be described with reference to FIG. In S101, the control unit 60 causes the component imaging camera 36 to capture an image of the bottom surface of the nozzle 41C.

[0085] In S102, the control unit 60 analyzes the captured image and estimates the degree of the following accumulation-type abnormality. Dirt-surface-electrode (accumulation type) Dirt - Surface - "Dust, lint" (accumulation type) Damage to property / wear-surface-cracks (accumulation type)

[0086] In S103, the control unit 60 analyzes the captured image to determine the presence or absence of the following sudden abnormalities. Dirt - Surface - Solder (accidental type) Damage to property / wear-surface-chip (sudden type) Damage to property / wear and tear - Other - Breakage (sudden type)

[0087] In S104, the control unit 60 measures the negative pressure when the nozzle 41C is pressed against the component E and the negative pressure when the nozzle 41C is not pressed against the component E, and estimates the difference between the measured negative pressures as the degree of abnormality for the following accumulation-type abnormalities. Dirt - Holes (clogged holes) - "Dust, lint" (accumulation type) Here, it is not possible to determine whether the cause of the clogged hole is "dust, lint" or embossed pieces based on the negative pressure difference alone. Therefore, the control unit 60 analyzes the image captured in S101 to determine whether the cause of the hole (clogged hole) is "dust, lint" or embossed pieces, and if it is "dust, lint," executes S104 to estimate the degree of the abnormality.

[0088] In S105, the control unit 60 analyzes the captured image to determine the presence or absence of the following sudden abnormalities. - Dirt - Holes (clogged holes) - Embossed pieces (sudden type) In S106, the control unit 60 lowers the mounting head 41 to press the nozzle 41C against the component E, then raises the mounting head 41, measures the negative pressure a predetermined time after the mounting head 41 started to rise, and estimates the measured negative pressure as the degree of the accumulation-type abnormality described below. Damage / wear - Other - Spring / grease deterioration (accumulation type)

[0089] (5-3-2) Estimation of the type of abnormality The process of estimating the type of abnormality is the same whether the degree of an individual abnormality is estimated as the degree of the abnormality or the degree of a common abnormality is estimated. A case will be described where the degree of an individual abnormality is estimated as the degree of the abnormality. In estimating the type of abnormality, the control unit 60 estimates the type of abnormality causing a pickup error (or the type of abnormality having a short remaining life until a pickup error occurs) or the type of abnormality causing a mounting failure (or the type of abnormality having a short remaining life until a mounting failure occurs) by comparing the degree of the individual abnormality for each type of abnormality with a first threshold value. In other words, the types of abnormality estimated include not only types of abnormality causing a pickup error or mounting failure, but also types of abnormality having a short remaining life. A type of abnormality having a short remaining life can also be said to be a type of abnormality about to occur.

[0090] The above-mentioned first threshold value may be different for each type of abnormality, or may be common to all types of abnormality. Specifically, as described above, the scale of the degree of abnormality differs depending on the type of abnormality. Therefore, a different first threshold value may be used for each type of abnormality. In contrast, for example, the degree of abnormality for each type of abnormality may be normalized to 0% (no abnormality) to 100% (abnormality present), and a first threshold value common to all types of abnormality may be used.

[0091] When estimating the degree of the common abnormality in the first estimation process described above, the control unit 60 estimates the type of abnormality using the degree of each individual abnormality. The reason for this is that since there is only one degree of common abnormality for one nozzle 41C, it is not possible to determine which type of abnormality has occurred and which has not occurred based on the degree of common abnormality.

[0092] (5-4) Judgment processing As described above, the determination process is a process for determining whether or not to predict the remaining life of the nozzle 41C for the type of abnormality estimated in the first estimation process. The control unit 60 determines whether or not to predict the remaining life for the type of abnormality estimated in the first estimation process by comparing the degree of the abnormality with a second threshold value (an example of a threshold value). Determination patterns in the determination process will be described below. Four patterns will be described here.

[0093] Pattern 1: Common degree of abnormality / Common second threshold In pattern 1, the control unit 60 judges whether to predict the remaining lifespan by comparing the degree of the common abnormality estimated in the first estimation process with a second threshold value common to all types of abnormality. In pattern 1, whether to predict the remaining lifespan is judged collectively for all types of abnormality estimated in the first estimation process. That is, if it is judged in the judgment process that the remaining lifespan is to be predicted, the remaining lifespan is predicted for all types of abnormality estimated in the first estimation process, and if it is judged that the remaining lifespan is not to be predicted, the remaining lifespan is not predicted for any type of abnormality.

[0094] Pattern 2: Common abnormality level / individual second threshold In pattern 2, the second threshold value differs depending on the type of abnormality. The control unit 60 judges whether to predict the remaining life by comparing the degree of the common abnormality estimated in the first estimation process with the second threshold value for each type of abnormality. In pattern 2, whether to predict the remaining life is judged for each type of abnormality.

[0095] Pattern 3: Individual abnormality level / common second threshold In pattern 3, the control unit 60 determines whether to predict the remaining life by comparing the degree of each abnormality estimated in the first estimation process with a second threshold common to all types of abnormality. For example, the control unit 60 normalizes the degree of each abnormality to 0% to 100%, and compares the normalized degree of abnormality with the common second threshold. In pattern 3 as well, whether to predict the remaining life is determined for each type of abnormality.

[0096] In pattern 3, the second threshold may be the same as the first threshold. When the first threshold and the second threshold are the same, the first estimation process and the judgment process may be integrated into one process. Specifically, when the normalized degree of abnormality is compared with the common first threshold (=common second threshold) in the first estimation process, the comparison of the individual degrees of abnormality with the common second threshold has already been performed in the first estimation process, so that it may be determined that the remaining life is to be predicted unconditionally for the type of abnormality estimated in the first estimation process.

[0097] Pattern 4: Individual degree of abnormality / individual second threshold In pattern 4, the second threshold value differs depending on the type of abnormality. The control unit 60 determines whether to predict the remaining life span by comparing the degree of each individual abnormality estimated in the first estimation process with the second threshold value for each type of abnormality. In pattern 4 as well, whether to predict the remaining life span is determined for each type of abnormality. In pattern 4 as well, the second threshold value may be the same as the first threshold value. If the first threshold value and the second threshold value are the same, the first estimation process and the judgment process can be combined into a single process.

[0098] (5-5) Prediction processing As described above, the control unit 60 predicts the remaining life of the nozzle 41C based on the remaining life evaluation value. The remaining life evaluation value may be the degree of the individual abnormality estimated in the first estimation process, or may be a normalized value of the degree of the individual abnormality. Apart from the degree of the individual abnormality estimated in the first estimation process, an evaluation value for predicting the remaining life may be estimated based on sensor information, and the estimated evaluation value may be used as the remaining life evaluation value.

[0099] For example, in the case of dirt-hole (clogged hole)-"dust, lint", if the degree of abnormality estimated in the first estimation process is used as the remaining life evaluation value, the difference between the negative pressure when the mounting head 41 is lowered to press the nozzle 41C against the component E and the negative pressure when the nozzle 41C is not pressed against the component E is used. Separately from this, the ratio (%) of the high abnormality area to the nozzle hole opening area may be estimated as the evaluation value. The high abnormality area is the area of ​​the region considered to be clogged.

[0100] The degree of each individual abnormality estimated in the first estimation process is used to determine whether or not to predict the remaining life, and high accuracy is not required as long as it is possible to determine whether or not to predict the remaining life. Since remaining life prediction is not necessarily performed for all types of abnormality, it takes time to estimate the degree of each individual abnormality with high accuracy for all types of abnormality. Therefore, by estimating the degree of each individual abnormality with relatively low accuracy without using too much information, the time required for estimation can be shortened.

[0101] In contrast, a certain degree of accuracy is required for the remaining life evaluation value in predicting the remaining life. For this reason, for the type of anomaly determined in the determination process to predict the remaining life, a highly accurate evaluation value may be estimated based on sensor information, and the evaluation value may be used as the remaining life evaluation value. Since the types of anomaly for which an evaluation value is estimated are only those types of anomaly determined in the determination process to predict the remaining life, a highly accurate evaluation value can be estimated in a relatively short time even if a large amount of information is used.

[0102] When the degree of each individual abnormality can be estimated with a relatively high degree of accuracy, the degree of each individual abnormality may be used as the remaining life evaluation value.

[0103] In Fig. 8, the upper graph (graph of the degree of abnormality) is a graph showing the relationship between the number of shots and the remaining life evaluation value, with the horizontal axis showing the number of shots and the vertical axis showing the remaining life evaluation value. One shot refers to one operation of picking up a component E and mounting it on a board P. In Fig. 8, the lower graph (graph of the incidence of mounting defects) is a graph showing the relationship between the number of shots and the incidence of mounting defects, with the horizontal axis showing the number of shots and the vertical axis showing the incidence of mounting defects.

[0104] In the graph shown in FIG. 8, when the remaining life evaluation value is the third threshold value Se, a pickup error or mounting failure begins to occur. Therefore, the third threshold value Se is a value at which a pickup error or mounting failure caused by the nozzle 41C begins to occur. As described above, the second threshold value Sth is a threshold value for determining whether or not to predict the remaining life. The control unit 60 predicts the remaining life for the types of abnormality whose remaining life evaluation value is equal to or greater than the second threshold value Sth. Here, among the types of abnormality whose remaining life evaluation value is equal to or greater than the second threshold value Sth, the types of abnormality whose remaining life evaluation value is less than the third threshold value Se are types of abnormality in which a pickup error or mounting failure has not yet occurred but is about to occur. The types of abnormality whose remaining life evaluation value is equal to or greater than the third threshold value Se are types of abnormality in which a pickup error or mounting failure has already occurred.

[0105] The second threshold value Sth will be described in more detail below, taking dirt-holes (clogged holes)-"dust, lint" as an example. First, a case will be described in which the degree of individual abnormality estimated in the first estimation process is used as a remaining life evaluation value. Before shipping the component mounting device 3, the developer of the component mounting device 3 uses a normal nozzle 41C to measure the difference in negative pressure between when the nozzle 41C is pressed against the component E and when it is not pressed against the component E multiple times, and determines the second threshold value Sth by the following formula 1. In formula 1, m is the average value of the difference in negative pressure in the normal nozzle 41C, σ is the standard deviation, and k is a safety factor. Second threshold Sth=|m+kσ| Equation 1

[0106] Next, a case where an evaluation value estimated separately from the degree of individual abnormalities is used will be described. In the case of dirt-holes (hole clogging)-"dust, lint", the ratio (%) of the highly abnormal area to the nozzle hole opening area can also be used as the evaluation value. The highly abnormal area is the area of ​​the region considered to be clogged. In this case, the second threshold value Sth is determined experimentally. For example, the ratio just before the ratio at which the suction force starts to decrease when the nozzle hole is blocked may be used as the second threshold value Sth.

[0107] Here, even when the degree of each abnormality estimated in the first estimation process is used as the remaining life evaluation value, the second threshold value Sth can be determined by the above-mentioned method. For example, the second threshold value Sth may be a negative pressure difference just before the negative pressure difference at which the suction force starts to decrease when the nozzle holes are blocked.

[0108] Although the example described here is dirt-hole (clogged hole)-"dust, lint," the evaluation value and the second threshold value Sth can be determined in a similar manner for other types of abnormality.

[0109] An example of remaining life prediction will be described below. Here, a case will be described in which the degree of abnormality estimated in the first estimation process is used as a remaining life evaluation value. For example, in the case of dirt - hole (clogged hole) - "dust, lint", a mounting failure will occur when the negative pressure difference (degree of abnormality) reaches 30 Pa (Pascals). Also, from past data, it is assumed that the number of shots required to reduce the negative pressure difference by 10 Pa is 100 shots. In this case, if the current negative pressure difference is 40 Pa, the number of shots remaining until the negative pressure difference reaches 30 Pa is 100 shots. Therefore, the remaining life is predicted to be 100 shots.

[0110] For example, in the case of physical damage / wear-surface-crack, assume that mounting failure will occur when the crack length (degree of abnormality) reaches 2 mm. Also, assume that past data shows that the number of shots (growth rate) required for the crack to grow 1 mm is 100 shots. In this case, if the current crack length is 0.5 mm, the remaining number of shots until the crack grows to 2 mm is 150 shots. Therefore, the remaining life is predicted to be 150 shots.

[0111] (5-6) Nozzle health check flow The flow of the nozzle health check executed by the control unit 60 will be described with reference to FIG.

[0112] In S201, the control unit 60 executes the first estimation process described above to estimate the degree and type of the abnormality (first estimation process). However, in the case of a sudden abnormality, the control unit 60 estimates only the presence or absence of the abnormality, and does not estimate the degree of the abnormality. In S202, the control unit 60 judges whether or not to predict the remaining life (judgment process), and if so, proceeds to S203, and if not, ends the process. In S203, the control unit 60 executes a prediction process and a notification process.

[0113] The flow of the prediction process and notification process executed in S203 will be described with reference to Fig. 10. For convenience, the following three types of abnormality will be described as examples. Damage / wear-surface-cracks Dirt - Surface - Solder Dirt - Holes (clogged holes) - "Dust, lint"

[0114] In S301, the control unit 60 judges whether the type of abnormality is a crack (material damage / wear-surface-crack) or not, and if it is a crack, the process proceeds to S302, and if it is not a crack, the process proceeds to S304. In S302, the control unit 60 predicts the remaining life (i.e., the number of remaining shots) from the current crack length and the past crack growth rate.

[0115] In S303, the control unit 60 issues a weak warning to the operator via the operation panel 61. A weak warning is a warning that notifies the user that an abnormality is about to occur in the nozzle 41C, but does not notify the user that immediate maintenance is required. Specifically, for example, the control unit 60 displays information indicating that a weak warning has been issued on the main screen of the operation panel 61, and displays the contents of the weak warning on a sub-screen different from the main screen. The contents of the weak warning include the predicted remaining life of the nozzle 41C, a message such as "Please perform nozzle maintenance soon," etc. In the weak warning, the control unit 60 may also display a maintenance method according to the type of abnormality.

[0116] In S304, the control unit 60 judges whether the type of abnormality is solder (dirt-surface-solder) or not, and if it is solder, proceeds to S305, and if it is not solder, proceeds to S307. Here, solder is a sudden abnormality. As described above, in the case of a sudden abnormality, the remaining life is judged to be 0 at the time when it is judged that the abnormality has occurred. Although the details will be described later, the control unit 60 issues a strong warning to the operator when the predicted remaining life is equal to or less than the set value. The strong warning is an example of notifying the operator that an abnormality that may affect a pickup error or a mounting failure has occurred in the nozzle 41C. In the first embodiment, the set value is set to 0. Therefore, in the case of a sudden abnormality, it is determined that the remaining life is equal to or less than the set value, and a strong warning, which will be described later, is issued. The set value may be a value greater than 0. When the set value is greater than 0, a strong warning is issued if the remaining life is equal to or less than the set value even for an accumulation type abnormality.

[0117] In S305, the control unit 60 displays an image of the bottom surface of the nozzle 41C on the operation panel 61, and instructs the operator to check the image of the nozzle 41C. In S306, the control unit 60 notifies the operator that an abnormality has occurred in the nozzle 41C by issuing a strong warning to the operator via the operation panel 61. Specifically, for example, the control unit 60 automatically retreats the nozzle 41C in which the abnormality has occurred to the nozzle station 38, and then displays a maintenance instruction for the nozzle 41C on the main screen. This makes it possible to continue automatic operation while maintaining mounting accuracy. Here, if the number of nozzles 41C is reduced, the production time per board (so-called cycle time) may become longer, and some operators may dislike a longer cycle time than the occurrence of minor pickup errors or mounting defects. For this reason, the control unit 60 may display a screen for the operator to permit the retraction of the nozzles 41C, and may retract the nozzles 41C if the operator permits.

[0118] As described above, the control unit 60 uses weak and strong warnings depending on the remaining life. In other words, the control unit 60 changes the notification method depending on the remaining life. The control unit 60 may also display a maintenance method according to the type of abnormality.

[0119] In S307, the control unit 60 judges whether the type of abnormality is a hole (hole clogging). If the type of abnormality is a hole (hole clogging), the control unit 60 proceeds to S308. In FIG. 10, the process for the case where the abnormality is not a hole (hole clogging) is omitted. In S308, the control unit 60 predicts the remaining life (i.e., the number of remaining shots) from the negative pressure difference and the rate of progress of the negative pressure difference in the past. In S309, the control unit 60 issues a weak warning on the display unit 72.

[0120] (6) Effects of the embodiment According to the component mounting device 3 (nozzle maintenance support device) of the first embodiment, the remaining life of the nozzle 41C until a pickup error or mounting failure caused by the nozzle 41C occurs is predicted, so that it is possible to notify the operator of the predicted remaining life, or to notify the operator of the remaining life itself but to give a guide such as "The nozzle will soon reach the end of its life. Please perform maintenance." This enables the operator to perform maintenance on the nozzle 41C before an abnormality occurs in the nozzle 41C. By providing the above-mentioned notification, even if a long time has passed since the last maintenance, the operator will not perform maintenance unless notified of the remaining life or maintenance guidance, and therefore unnecessary maintenance of the nozzle 41C can be suppressed compared to when maintenance is performed based on time management. This makes it possible to minimize the frequency of maintenance. Therefore, according to the component mounting apparatus 3, it is possible to perform maintenance on the nozzle 41C before an abnormality occurs in the nozzle 41C, while minimizing the frequency of the maintenance. According to the component mounting apparatus 3, the remaining life is predicted for each type of abnormality of the nozzle 41C, so that the remaining life can be predicted more accurately than when the remaining life is predicted in a state where various types of abnormalities are mixed.

[0121] According to the component mounting apparatus 3, the remaining life of the nozzle 41C is notified to the operator, so that the operator can determine whether or not maintenance of the nozzle 41C is required. Therefore, unnecessary maintenance of the nozzle 41C can be suppressed compared to the case where maintenance is performed based on time management.

[0122] According to the component mounting device 3, for example, when the remaining life is short (for example, when the remaining life is equal to or less than a set value), the remaining life of the nozzle 41C is notified to the operator in a manner that reliably notifies the operator (i.e., a strong warning), which increases the likelihood that the operator will immediately perform maintenance on the nozzle 41C. On the other hand, when the remaining life is long, the notification is made in a manner that is less likely to interfere with the operator's work (i.e., a weak warning), which prevents the notification of the remaining life from interfering with the operator's work.

[0123] According to the component mounting device 3, the maintenance method according to the type of abnormality is notified, so for example, in the case of a type of abnormality that can be maintained by an automatic cleaning machine, the maintenance method using the automatic cleaning machine is notified as the maintenance method, making it possible to effectively use the automatic cleaning machine and reducing work requiring manpower as much as possible. This makes it possible to perform maintenance with the minimum amount of effort required.

[0124] <Embodiment 2> In the second embodiment, the management device 2 (an example of a computer) functions as a nozzle maintenance support device. A nozzle maintenance support program is stored in a memory unit 71 of the management device 2 according to the second embodiment. In the second embodiment, the communication unit 74 provided in the management device 2 is an example of an acquisition unit, and the display unit 72 is an example of a notification unit. The management device 2 acquires the sensor information by receiving the sensor information from the component mounting device 3 via the communication unit 74.

[0125] 11, a sequence chart will be described in which the management device 2 receives sensor information from the component mounting device 3 and stores it in the storage unit 71. The component mounting device 3 acquires the sensor information (S401), and transmits the acquired sensor information and nozzle identification information to the management device 2 (S402). The management device 2 stores the sensor information and nozzle identification information received from the component mounting device 3 in the storage unit 71 (S403).

[0126] After storing the sensor information and the nozzle identification information in the memory unit 71, the management device 2 may predict the remaining life immediately, or, rather than making an immediate prediction, may make the prediction when the nozzle maintenance support program is started by scheduled startup or timer startup. In other respects, the second embodiment is substantially the same as the first embodiment.

[0127] According to the management device 2 (nozzle maintenance support device) of embodiment 2, the management device 2 which is communicatively connected to the component mounting device 3 functions as a nozzle maintenance support device, and the management device 2 receives information from the component mounting device 3, thereby making it possible to predict the remaining life span without providing each component mounting device 3 with a control unit 60 with high processing power.

[0128] According to the management device 2, by making a computer (management device 2) separate from the component mounting device 3 function as a nozzle maintenance support device, it is possible to suppress a decrease in the production efficiency of the component mounting device 3.

[0129] <Embodiment 3> The third embodiment is a modification of the second embodiment. The processing unit 70 according to the third embodiment receives additional information for estimating an evaluation value for predicting the remaining life for the type of abnormality determined in the determination process to be the one for predicting the remaining life from the component mounting device 3 via the communication unit 74 (an example of an acquisition process), and estimates the evaluation value based on the received additional information (an example of a second estimation process). Then, the processing unit 70 predicts the remaining life based on the estimated evaluation value.

[0130] As shown in Fig. 12, component mounting apparatus 3 according to embodiment 3 includes a side-view camera 80. Side-view camera 80 is provided in head unit 34. Side-view camera 80 includes light source 80A arranged to the left of nozzle 41C in Fig. 12, and imaging unit 80B arranged to the right of nozzle 41C. Light emitted from light source 80A that is not blocked by nozzle 41C (transmitted light) enters imaging unit 80B, whereby a side-view image showing the shadow of nozzle 41C is captured. Component mounting apparatus 3 transmits the side-view image to management device 2 as additional information.

[0131] The flow of the nozzle health check according to the third embodiment will be described with reference to Fig. 13. Here, the same steps as those in the nozzle health check according to the first embodiment are denoted by the same reference numerals, and the description thereof will be omitted.

[0132] In S501, the processing unit 70 estimates an evaluation value based on additional information (an example of a second estimation process).

[0133] The second estimation process executed in S501 will be described with reference to Figures 14A, 14B, and 14C. Here, the following examples are used to explain the second estimation process: Damage / wear-surface-chipped, dirt-surface-solder, and dirt-hole (clogged hole)-"dust, lint."

[0134] (1) Damage / wear - surface - chipping The flow in the case of physical damage / wear-surface-chip will be described with reference to FIG. 14A. In S601, processing unit 70 requests component mounting apparatus 3 to transmit side-view images. Component mounting apparatus 3 captures multiple side-view images of nozzle 41C with side-view camera 80 while rotating mounting head 41 around its axis by a predetermined angle at a time, and transmits the captured multiple side-view images to management device 2.

[0135] In S602, the processing unit 70 analyzes the received side view image and determines the size of the chipped nozzle 41C. In S603, the processing unit 70 estimates an evaluation value based on the determined size of the chipping.

[0136] (2) Dirt - Surface - Solder The flow for the stain-surface-solder case will be described with reference to FIG. 14B. In S701, the processing unit 70 requests the component mounting device 3 to transmit an image of the underside of the nozzle 41C captured by the component imaging camera 36 under lighting conditions that emphasize specular reflection. The component mounting device 3 captures an image of the underside of the nozzle 41C under lighting conditions that emphasize specular reflection, and transmits the captured image to the management device 2.

[0137] In S702, the processing unit 70 analyzes the received image and checks the brightness of the abnormal area. In S703, the processing unit 70 judges whether or not solder is attached to the lower surface of the nozzle 41C based on the check result in S702, in other words, judges whether or not an abnormality of the type of dirt-surface-solder has occurred.

[0138] (3) Dirt - Holes (clogged holes) - "Dust, lint" The flow in the case of dirt-hole (clogged hole)-"dust, lint" will be described with reference to FIG. 14C. In S801, the processing unit 70 instructs the operator to perform air blowing. Air blowing is a task in which the operator removes the nozzle 41C and uses an external air blowing device to blow away dust and lint inside the nozzle hole with high-pressure air. Then, the processing unit 70 requests the component mounting device 3 to send the negative pressure difference after the air blowing is completed. The component mounting device 3 measures the negative pressure difference again and transmits it to the management device 2. In S802, the processing unit 70 estimates an evaluation value of the hole (hole clogging) based on the received negative pressure difference.

[0139] (4) Effects of the embodiment According to the management device 2 (nozzle maintenance support device) of the third embodiment, it is usually possible to determine with a small amount of processing whether to predict the remaining life by estimating the degree of abnormality with a small amount of information. Then, according to the nozzle maintenance support device, for the type of abnormality for which it has been determined that the remaining life should be predicted, additional information is obtained from the component mounting device 3 and an evaluation value is estimated, thereby making it possible to accurately predict the remaining life.

[0140] <Embodiment 4> The nozzle maintenance support device according to the fourth embodiment is a management device 2. The processing unit 70 according to the fourth embodiment stores the estimated remaining life evaluation value in the storage unit 71 (an example of a first storage process), and estimates a current remaining life evaluation value based on past remaining life evaluation values ​​stored in the storage unit 71 (an example of an estimation process). If the difference between the current estimated remaining life evaluation value and the current remaining life evaluation value estimated in the estimation process is equal to or greater than a reference value, the processing unit 70 reacquires sensor information and re-estimates the current remaining life evaluation value (an example of a re-estimation process), and in the prediction process, predicts the remaining life of the nozzle 41C based on the remaining life evaluation value re-estimated in the re-estimation process.

[0141] When the difference is less than the reference value, the processing unit 70 corrects the currently estimated remaining life evaluation value based on the past remaining life evaluation value stored in the storage unit 71 (an example of a correction process). Then, in a prediction process, the processing unit 70 predicts the remaining life of the nozzle 41C based on the remaining life evaluation value corrected in the correction process.

[0142] The date and time when the operator performed the maintenance is also stored in the storage unit 71. The date and time when the maintenance was performed may be input by the operator. Alternatively, when the automatic cleaning machine 4 is connected to the communication network 5 and the operator uses the automatic cleaning machine 4 to clean the nozzle 41C, the automatic cleaning machine 4 may transmit the date and time when the maintenance of the nozzle 41C was performed to the management device 2, and the management device 2 may store it in the storage unit 71. An example of re-estimation and correction of the remaining life evaluation value will be described below.

[0143] (1) Re-estimation example An example of re-estimation will be described with reference to Fig. 15. In Fig. 15, time point T is the time point when the last maintenance was performed on nozzle 41C. The remaining life evaluation value is large immediately before maintenance, but after maintenance is performed, the abnormality is eliminated and the remaining life evaluation value becomes small.

[0144] In Fig. 15, white circles 90 are past actual remaining life evaluation values, and black circles 91 are currently estimated remaining life evaluation values. Of the past actual remaining life evaluation values, those that are equal to or greater than the second threshold value Sth are generally on line 92, so it is assumed that the currently estimated remaining life evaluation value (black circle 91) is also on line 92, but in the example shown in Fig. 15, it deviates significantly from line 92. In other words, it deviates from the actual value.

[0145] For this reason, the processing unit 70 obtains an approximation line 92 of the remaining life evaluation value from past actual values, and if the difference between the currently estimated remaining life evaluation value and the current remaining life evaluation value inferred from the approximation line 92 is equal to or greater than a reference value, it determines that the estimation may have failed. In that case, the processing unit 70 reacquires sensor information from the component mounting device 3 and re-estimates (i.e., retries) the remaining life evaluation value, and predicts the remaining life using the re-estimated remaining life evaluation value. In Fig. 15, the shaded circle 93 indicates the re-estimated remaining life evaluation value.

[0146] (2) Correction Example 1 Correction Example 1 will be described with reference to Fig. 16. If the difference between the currently estimated remaining life evaluation value (black circle 91) and the current remaining life evaluation value estimated from approximation line 92 (shaded circle 93) is less than a reference value, processing unit 70 does not re-estimate the remaining life evaluation value, and predicts the remaining life using the current remaining life evaluation value estimated from approximation line 92. That is, the currently estimated remaining life evaluation value is corrected with the current remaining life evaluation value estimated from approximation line 92.

[0147] However, the processing unit 70 stores the currently estimated remaining life evaluation value as an actual value, and in the next and subsequent remaining life predictions, the processing unit 70 obtains the approximation line 92 using the currently estimated remaining life evaluation value.

[0148] (3) Correction Example 2 Correction example 2 will now be described with reference to Fig. 17. In the above-described correction example 1, the remaining life evaluation value is corrected based on the actual value of the remaining life evaluation value after the last maintenance was performed. In contrast, in correction example 2, the actual value before the last maintenance is performed is also used to correct the remaining life evaluation value.

[0149] Here, one cycle is defined as the period from when the maintenance of nozzle 41C is performed to when the next maintenance is performed. In the example shown in Fig. 17, processing unit 70 uses the performance values ​​in the past cycle (or the past multiple cycles) in addition to the performance values ​​in the current cycle to obtain approximate straight line 92. The method of correction after obtaining approximate straight line 92 is the same as in correction example 1.

[0150] (4) Effects of the embodiment According to the management device 2 of embodiment 4, the current remaining life evaluation value is re-estimated, and by eliminating sudden outliers (failure to estimate the remaining life evaluation value), the remaining life evaluation value can be estimated more accurately (and thus the remaining life can be predicted more accurately).

[0151] According to the management device 2, the currently estimated remaining life evaluation value is corrected based on the previously estimated remaining life evaluation value, so that the accuracy of prediction of the remaining life is improved. The management device 2 makes it possible to eliminate the influence of minute estimation errors in the remaining life evaluation value and to estimate a stable remaining life evaluation value (and thus to predict the remaining life with stability). When the predicted remaining life is linked to a production plan, variations in the remaining life prediction accuracy may cause the production plan to change suddenly, causing chaos at the production site. A stable prediction of the remaining life can reduce the risk of chaos at the production site. The management device 2 enables estimation of the remaining life evaluation value (and thus stable prediction of the remaining life) by taking into account more actual results. In particular, it is possible to suppress variation in the remaining life evaluation value at the beginning of the rise of the remaining life evaluation value.

[0152] <Embodiment 5> The nozzle maintenance support device according to the fifth embodiment is a management device 2. In the fifth embodiment, the processing unit 70 estimates the degree and type of anomaly using a classifier created by machine learning (including deep learning). The classifier is a computer program executed by the processing unit 70. When estimating the degree of anomaly using a classifier, the degree of an individual anomaly may be estimated, or the degree of a common anomaly may be estimated. In the fifth embodiment, an example will be described in which the degree of a common anomaly is estimated.

[0153] In the fifth embodiment, it is assumed that the classifier is created by the manufacturer of the component mounting apparatus 3. The manufacturer of the component mounting apparatus 3 creates the classifier by machine learning and provides the created classifier to the user of the component mounting apparatus 3. For this reason, learning is not performed on the user side. Specifically, the classifier is created using the component mounting apparatus 3 and a learning computer (hereinafter referred to as the learning computer) provided by the manufacturer of the component mounting apparatus 3.

[0154] (1) Learning Method Here, two methods are explained as examples of machine learning. Either of the following two methods may be used.

[0155] (1-1) Method 1 In method 1, the degree and type of abnormality are estimated in the following two steps. [Step 1] The learning computer learns a model capable of estimating the degree of common abnormality through unsupervised learning (PatchCore, PaDiM, AnoGAN, etc.). Specifically, the learning computer learns the normal state from an image of the underside of the normal nozzle 41C captured by the part imaging camera 36, ​​and learns the above-mentioned model from the degree of deviation from the normal state. In method 1, the model is learned from the degree of deviation from the normal state, so the learning computer can learn the model with a small number of abnormal images. [Step 2] The training computer learns the types of anomalies using a clustering technique. Clustering is a technique that compares the features of images and categorizes similar images into the same class. Because the training computer learns the types of anomalies using a clustering technique, it can learn the types of anomalies using a small number of abnormal images.

[0156] (1-2) Method 2 In method 2, the degree and type of abnormality are estimated in the following one step. [Step 1] The training computer uses techniques such as SSD (Single Shot Multi-Box Detection) and YOLO to simultaneously estimate the degree of anomaly and, if an anomaly exists, the type of anomaly, as a so-called object detection problem. When making simultaneous estimates, abnormal images for each type of anomaly are also used. The training computer estimates the type of anomaly based on which abnormal image it is similar to. Method 2 can estimate the degree and type of anomaly in one step, but it requires many abnormal images for each type of anomaly.

[0157] (2) Judgment processing As described above, the degree of the common abnormality is estimated by the classifier in the embodiment 5. When the degree of the common abnormality is estimated, the processing unit 70 judges whether or not to predict the remaining life in the judgment process based on the above-mentioned pattern 1 (degree of common abnormality / common second threshold value) or pattern 2 (degree of common abnormality / individual second threshold value). However, even when the degree of a common abnormality is estimated, it may be judged using pattern 3 (degree of individual abnormality / common second threshold) or pattern 4 (degree of individual abnormality / individual second threshold). For example, if the trend of the degree of common abnormality matches the trend of the degree of individual abnormality, it may be judged using pattern 3 or pattern 4. As described above, the degree of individual abnormalities may be estimated using a classifier. In that case, the determination is made according to pattern 3 or pattern 4.

[0158] (3) Prediction processing Since the remaining life is predicted for each type of anomaly, a remaining life evaluation value is required for each type of anomaly in order to predict the remaining life. When estimating the degree of a common anomaly, the degree of each individual anomaly is not estimated, so the degree of each individual anomaly cannot be used as the remaining life evaluation value. For this reason, when estimating the degree of a common anomaly, an evaluation value is estimated for each type of anomaly based on sensor information, separate from the degree of the common anomaly. Then, this evaluation value is used as the remaining life evaluation value for predicting the remaining life.

[0159] (4) Effects of the embodiment According to the management device 2 (nozzle maintenance support device) of the fifth embodiment, the classifier is created by machine learning, so that the classifier can be created with little effort. The management device 2 learns the degree of abnormality through unsupervised learning. There are various types of nozzles, differing in the size, shape, and number of nozzle holes. When creating a classifier through supervised learning using training data, if an attempt is made to learn the degree of abnormality for each type of abnormality for all nozzle types, training data covering all types of abnormality and all degrees of abnormality for all nozzle types must be prepared, which is difficult to achieve. Learning the degree of abnormality through unsupervised learning makes it possible to learn the degree of abnormality with little effort.

[0160] According to the management device 2, different types of nozzles 41C can be supported by changing the normal state image used for unsupervised learning. Therefore, even if there are multiple types of nozzles 41C, a classifier can be created in a short period of time.

[0161] According to the management device 2, since an abnormality is defined using an image of a normal state, it may be possible to detect an abnormality that is difficult to define logically.

[0162] According to the management device 2, the classifier is created by the manufacturer of the component mounting device 3. Therefore, by having an expert from the manufacturer of the component mounting device 3 carry out the learning, which requires a computer with high processing power and know-how, in advance, the user can accurately estimate the type and degree of the abnormality even with a computer with low processing power.

[0163] <Embodiment 6> The sixth embodiment is a modification of the fifth embodiment. In the sixth embodiment, a classifier is also created by the manufacturer of the component mounting apparatus 3, but the created classifier is re-trained on the user side. Specifically, in the sixth embodiment, previously acquired sensor information, the degree of anomaly and the type of anomaly estimated by the classifier based on the sensor information are associated with each other and stored in the storage unit 71 of the management device 2. The operator checks the degree of anomaly and the type of anomaly stored in the storage unit 71 at an appropriate timing, and corrects them via the operation unit 73 if the stored degree of anomaly or type of anomaly is not appropriate. The process in which the management device 2 accepts the corrections from the operator via the operation unit 73 is an example of a correction acceptance process. The management device 2 re-learns the classifier based on the correction results accepted in the correction acceptance process (an example of a re-learning process).

[0164] With reference to Fig. 18, a correction screen 100 for an operator to correct the degree and type of abnormality will be described. A graph showing the estimated degree of abnormality is displayed on the correction screen 100. The horizontal axis of the graph is the number of shots, and the vertical axis is the degree of abnormality. In the graph, the type of abnormality is represented by the color of the circles 101 or the type of hatching. When the operator operates the operation panel 61 to select one of the circles 101, a pop-up screen 102 is displayed, and detailed information on the sensor information and the estimation result corresponding to the selected circle 101 is displayed.

[0165] Specifically, in the example shown in Fig. 18, information such as an image of the underside of nozzle 41C, the estimated type of abnormality (dust clogging), and the degree of abnormality (50%) are displayed. The operator looks at the displayed image and judges whether the degree and type of abnormality are appropriate, and corrects them if they are not appropriate. In the sixth embodiment, the type of abnormality can be corrected by selecting it from a list. The degree of abnormality is to be directly entered into a text box.

[0166] After correcting them, the operator presses the update button 103. When the operator presses the update button 103, the estimation result corresponding to the selected circle 101 is updated.

[0167] (1) Classifier re-learning process flow 19, a flow of the classifier re-learning process executed by the processing unit 70 of the management device 2 will be described. This process is executed periodically. In S1001, the processing unit 70 re-learns the classifier based on the past sensor information and the estimation result stored in the storage unit 71.

[0168] In S1002, the processing unit 70 inputs the past sensor information stored in the memory unit 71 to the classifier (the re-learned classifier) ​​to re-estimate the degree and type of abnormality. Then, the processing unit 70 compares the re-estimated estimation result with the past estimation results stored in the memory unit 71 to determine the matching rate therebetween. For example, it is assumed that the component mounting device 3 estimates the degree and type of abnormality 100 times and the sensor information and estimation results for the 100 times are stored in the memory unit 71. In this case, the processing unit 70 performs re-estimation using the sensor information for the 100 times stored in the memory unit 71 and determines the rate at which the estimation results match. For example, if the estimation results match 80 times out of 100 times, the matching rate is 80%.

[0169] In S1003, if the matching rate is less than a predetermined reference value, the processing unit 70 determines that the difference in the estimation results is large and proceeds to S1004, whereas if the matching rate is equal to or greater than the reference value, the processing unit 70 determines that the difference is small and ends the process. In S1004, the processing unit 70 displays a query screen (see FIG. 20) to be described later and queries the operator as to whether or not to update the classifier. If the operator selects to update, the processing unit 70 proceeds to S1005, and if the operator selects not to update, the processing unit 70 terminates the process. In S1005, the processing unit 70 updates the classifier.

[0170] 20, the inquiry screen 110 displayed in S1004 will be described. A graph of the degree of abnormality previously estimated and stored in the storage unit 71 and a graph of the degree of abnormality estimated by the recreated classifier are displayed on the inquiry screen 110. The above-mentioned matching rate may also be displayed on the inquiry screen 110.

[0171] (2) Effects of the embodiment According to the maintenance support device of the sixth embodiment, if the estimation result of the classifier is incorrect, the operator corrects the estimation result and the classifier is re-learned using the corrected result, thereby improving the reliability of the estimation result. Here, there is a possibility that the maintenance support device will fail to re-learn in S1001. According to the maintenance support device, in S1004, a query screen 110 is displayed to query the operator as to whether or not to update the classifier, and if the operator selects not to update, the classifier is not updated, so that it is possible to prevent the classifier that failed in the re-learning from being used.

[0172] <Embodiment 7> The nozzle maintenance support device according to the seventh embodiment is a management device 2. A seventh embodiment will be described with reference to Fig. 21. In the seventh embodiment, the processing unit 70 of the management device 2 learns a third threshold value Se for determining that the remaining life is 0 (in other words, a remaining life evaluation value for determining that the remaining life is 0) based on the remaining life evaluation value at the time when maintenance was performed on the nozzle 41C before an abnormality occurred in the nozzle 41C (an example of a learning process).

[0173] Specifically, when an operator performs maintenance on nozzle 41C, the operator inputs the date and time (date and time when maintenance was performed) to management device 2. Processing unit 70 learns the remaining life evaluation value of nozzle 41C at the time when maintenance was performed as a remaining life evaluation value (third threshold value Se′) at which the remaining life is determined to be 0.

[0174] Here, when the automatic cleaning machine 4 is connected to the communication network 5 and an operator uses the automatic cleaning machine 4 to clean the nozzle 41C, the automatic cleaning machine 4 may transmit the date and time when maintenance of the nozzle 41C was performed to the management device 2 via the communication network 5.

[0175] (1) Learning process The flow of the above-mentioned learning process will be described with reference to FIG. In S1101, the processing unit 70 receives an input of the date and time of maintenance from the operator. In S1102, the processing unit 70 receives input of the reason for maintenance from the operator. The reason for maintenance refers to the type of abnormality that caused the operator to perform maintenance. For example, when maintenance is performed to deal with property damage / wear-surface-cracks, property damage / wear-surface-cracks is input as the reason for maintenance.

[0176] In S1103, the processing unit 70 judges whether the type of abnormality input as the reason for maintenance is an accumulation type or an unexpected type. If it is an accumulation type, the processing unit 70 proceeds to S1104, and if it is an unexpected type, the processing unit 70 ends the processing. The reason for ending the processing in the unexpected type is that the remaining life is not predicted in the unexpected type. In S1104, the processing unit 70 updates the third threshold Se with the remaining life evaluation value at the time when the maintenance was performed. In this way, the remaining life evaluation value (third threshold Se') at which the remaining life is determined to be 0 is learned.

[0177] (2) Effects of the embodiment According to the nozzle maintenance support device of the seventh embodiment, the remaining life evaluation value at which the remaining life is determined to be zero is learned, thereby improving the prediction accuracy of the remaining life.

[0178] <Embodiment 8> The nozzle maintenance support device according to the eighth embodiment is a management device 2. An eighth embodiment will be described with reference to Fig. 23. In the eighth embodiment, a management device 2, a plurality of component mounting devices 3, an automatic cleaner 4, and a data storage server 6 are connected to a communication network 5.

[0179] Each component mounting device 3 associates the sensor information with the nozzle identification information and transmits them to the management device 2, and the management device 2 estimates a remaining life evaluation value for each type of abnormality based on the received sensor information. Then, the management device 2 associates the nozzle identification information with the remaining life evaluation value and stores them in the data storage server 6 (an example of a second storage process). Then, when the management device 2 corrects the remaining life evaluation value currently estimated for a certain nozzle 41C, it reads out the remaining life evaluation value associated with the nozzle identification information of that nozzle 41C from the data storage server 6, and makes corrections using the read remaining life evaluation value.

[0180] The information stored in association with the nozzle identification information is not limited to the remaining life evaluation value, and other information may be stored instead of (or in addition to) the remaining life evaluation value. For example, sensor information may be stored, the estimated degree or type of abnormality may be stored, or maintenance information (such as the date and time when maintenance was performed) may be stored.

[0181] In the eighth embodiment, when maintenance of the nozzle 41C is performed using the automatic cleaning machine 4, the automatic cleaning machine 4 transmits the nozzle identification information and a maintenance record (the date and time when the maintenance was performed and the details of the maintenance) to the management device 2. The processing unit 70 associates the maintenance record received from the automatic cleaning machine 4 with the nozzle identification information and stores it in the data storage server 6. If the automatic cleaning machine 4 is not connected to the communication network 5, an operator may input the nozzle identification information and the maintenance record to the management device 2, and the management device 2 may store the inputted nozzle identification information and maintenance record in the data storage server 6.

[0182] According to the management device 2 of embodiment 8, the estimated remaining life evaluation value is stored in the data storage server 6, so that regardless of which nozzle the remaining life evaluation value is estimated by the maintenance support device, it becomes easy to use all remaining life evaluation values ​​previously estimated for a certain nozzle 41C for some kind of processing (for example, the re-estimation processing or correction processing described above).

[0183] In the eighth embodiment, the automatic cleaning machine 4 is connected to the communication network 5, so that the maintenance records can be acquired automatically. Therefore, the remaining life evaluation value can be corrected taking the maintenance records into consideration.

[0184] <Embodiment 9> In the above-described embodiment, the nozzle health check is performed periodically, but the nozzle health check may be performed when a specific event occurs instead of periodically. Examples of events for which a nozzle health check is performed are described below.

[0185] Event example 1: Shutdown of component mounting equipment 3 (for users whose component mounting systems are shut down at night). A nozzle health check is incorporated into the shutdown operation, and is performed when component mounting and other operations for the day are completed and the component mounting equipment 3 is shut down.

[0186] Event example 2: Number of shots (for users who perform continuous operation). For each nozzle 41C, a nozzle health check is performed when the number of shots reaches a predetermined number.

[0187] Event example 3: Prioritize and perform as much as possible (for users who value efficiency). Priorities are set in advance for each nozzle 41C (or for each type of nozzle 41C), and nozzle health checks are performed as much as possible in order from the nozzles 41C with the highest priority during free time such as during substrate transport.

[0188] Event example 4: When an abnormality occurs (mainly when a sudden abnormality occurs). When a mounting defect is discovered by the inspection machine or when pickup errors occur / increase, a nozzle health check is performed for the corresponding nozzle 41C.

[0189] <Other embodiments> The technology disclosed in this specification is not limited to the embodiments described above and in the drawings, and for example, the following embodiments are also included in the technical scope disclosed in this specification.

[0190] (1) In the above embodiment, when the remaining life is notified, a maintenance method according to the type of abnormality in the nozzle 41 C is also notified. However, it is also possible to notify only the remaining life, without notifying the maintenance method.

[0191] (2) In the above embodiments 4-8, the management device 2 is exemplified as the nozzle maintenance support device. However, the nozzle maintenance support device may be the component mounting device 3.

[0192] (3) In the above embodiment, the surface mounter 3 and the management device 2 are exemplified as examples of the nozzle maintenance support device, but the nozzle maintenance support device is not limited to these. For example, some automatic cleaners 4 in recent years are equipped with a function for inspecting the nozzle 41C, so the automatic cleaner 4 that has an inspection function may be given the function of the nozzle maintenance support device. [Explanation of symbols]

[0193] 2: Management device (an example of a nozzle maintenance support device and computer) 3: Component mounting device (an example of a nozzle maintenance support device) 36: Part imaging camera (an example of the acquisition unit) 41C: Nozzle 60: Control unit (an example of a processing unit) 61: Operation panel (an example of a notification section and an operation section) 70: Processing section 71: Memory unit P: Substrate E: Component

Claims

1. A maintenance support device for a component suction nozzle provided in a component mounting device that mounts components on a board, comprising: An acquisition unit that acquires information regarding the nozzle; A processing section; Equipped with The processing unit includes: a first estimation process for estimating a degree and a type of abnormality of the nozzle based on the information acquired by the acquisition unit; a determination process for determining whether or not to predict a remaining life of the nozzle by comparing a degree of the abnormality estimated in the first estimation process with a threshold value for the type of abnormality estimated in the first estimation process; a prediction process for predicting the remaining life of the nozzle until a pickup error or mounting failure caused by the nozzle occurs, for the type of abnormality determined to predict the remaining life in the determination process, based on the remaining life evaluation value, when the degree of abnormality estimated in the first estimation process is defined as a remaining life evaluation value, or when an evaluation value estimated based on the information acquired by the acquisition unit and for predicting the remaining life of the nozzle is defined as a remaining life evaluation value; To perform nozzle maintenance support equipment.

2. The nozzle maintenance support device according to claim 1, The processing unit executes a first notification process of notifying an operator of the remaining life predicted in the prediction process via a notification unit.

3. The nozzle maintenance support device according to claim 2, The processing unit, in the first notification process, varies a notification method depending on the remaining life predicted in the prediction process.

4. A nozzle maintenance support device according to any one of claims 1 to 3, The processing unit executes a second notification process that notifies an operator via a notification unit of a maintenance method corresponding to the type of abnormality for which the remaining life has been predicted in the prediction process.

5. A nozzle maintenance support device according to any one of claims 1 to 3, The processing unit executes a third notification process to notify an operator via a notification unit that an abnormality has occurred in the nozzle that may affect suction errors or mounting defects when the remaining life predicted in the prediction process is equal to or less than a set value.

6. A nozzle maintenance support device according to any one of claims 1 to 3, the nozzle maintenance support device is a computer connected to the component mounting device so as to be able to communicate with the component mounting device, The acquisition unit acquires the information by receiving the information from the component mounting device.

7. A nozzle maintenance support device according to any one of claims 1 to 3, The processing unit includes: an acquisition process of acquiring additional information for estimating the evaluation value via the acquisition unit; a second estimation process for estimating the evaluation value based on additional information acquired by the acquisition unit; Run A nozzle maintenance support device, wherein in the prediction process, a remaining life is predicted based on the evaluation value estimated in the second estimation process.

8. A nozzle maintenance support device according to any one of claims 1 to 3, The processing unit includes: a first storage process for storing the estimated remaining life evaluation value in a storage unit; an estimation process of estimating a current remaining life evaluation value based on the past remaining life evaluation values ​​stored in the storage unit; a re-estimation process for re-acquiring information about the nozzle and re-estimating the current remaining life evaluation value when a difference between the currently estimated remaining life evaluation value and the current remaining life evaluation value estimated in the estimation process is equal to or greater than a reference value; Run A nozzle maintenance support device, wherein in the prediction process, a remaining life of the nozzle is predicted based on the remaining life evaluation value re-estimated in the re-estimation process.

9. A nozzle maintenance support device according to any one of claims 1 to 3, The processing unit includes: a first storage process for storing the remaining life evaluation value in a storage unit; a correction process for correcting the currently estimated remaining life evaluation value based on the past remaining life evaluation value stored in the storage unit; Run A nozzle maintenance support device, wherein in the prediction process, a remaining life of the nozzle is predicted based on the remaining life evaluation value corrected in the correction process.

10. A nozzle maintenance support device according to any one of claims 1 to 3, The processing unit estimates the degree and type of abnormality using a classifier created by unsupervised learning.

11. The nozzle maintenance support device according to claim 10, Equipped with an operating unit, The processing unit includes: a correction receiving process of receiving, from an operator via the operation unit, a correction of at least one of the degree of abnormality and the type of abnormality estimated by the classifier; a re-learning process for re-learning the classifier based on the correction result accepted in the correction acceptance process; To perform nozzle maintenance support equipment.

12. A nozzle maintenance support device according to any one of claims 1 to 3, The processing unit executes a learning process to learn the remaining life evaluation value that determines that the remaining life is 0 based on the remaining life evaluation value at the time when maintenance is performed on the nozzle before a suction error or mounting defect caused by the nozzle occurs.

13. A nozzle maintenance support device according to any one of claims 1 to 3, The processing unit executes a second storage process of storing nozzle identification information of the nozzle and the remaining life evaluation value in association with each other in a data storage server.

14. A maintenance support program for a component suction nozzle provided in a component mounting device that mounts components on a board, comprising: a first estimation process for estimating a degree and a type of abnormality of the nozzle based on information about the nozzle; a determination process for determining whether or not to predict a remaining life of the nozzle by comparing a degree of the abnormality estimated in the first estimation process with a threshold value for the type of abnormality estimated in the first estimation process; a prediction process for predicting a remaining life of the nozzle until a pickup error or mounting failure caused by the nozzle occurs, for the type of abnormality determined in the determination process to predict a remaining life, based on the remaining life evaluation value, when the degree of abnormality estimated in the first estimation process is defined as a remaining life evaluation value, or when an evaluation value estimated based on information about the nozzle and for predicting a remaining life of the nozzle is defined as a remaining life evaluation value; A nozzle maintenance support program that causes a computer to execute the above.

Images