TW202108277A - Machine-learning system for laser processing device and machine-learning method - Google Patents

Abstract

In order to promote automation of a portion or the entirety of the operations of a plurality of laser processing devices, this machine-learning system for a laser processing device is configured which includes: a learning data storage unit 101 which stores acquired learning data that is a set of images captured by a camera sensor for capturing images of objects to be processed respectively provided in the plurality of laser processing devices that irradiate target positions with laser light in order to correct defects generated in the objects to be processed, and additional information that suggests, at least in a corresponding image, a defect position of the object to be processed or a target position to be irradiated with the laser light; and a learning unit 102 which generates, by using a learning data group stored in the learning data storage unit 101, a learnt model in which the image of the object to be processed, which is captured with the camera sensor provided in the laser processing device, is taken as an input and the defect position or the target position at least in the image is taken as an output.

Description

Machine learning system and machine learning method for laser processing device

The present invention relates to the control of a laser processing device that performs processing by irradiating laser light toward a target position on an object to be processed.

As a kind of laser processing device, there can be exemplified corrections in liquid crystal display modules, plasma display modules, electro-luminescence display modules, inorganic EL display modules, and micro LEDs (Light-Emitting Diode). ) Laser repair device for defects (or defects) caused by display modules, transparent conductive film substrates, or color filters (for example, refer to the following patent documents). [Prior Technical Literature] [Patent Literature]

[Patent Document 1] Japanese Patent Application Publication No. 2005-274709

[The problem to be solved by the invention]

In this kind of laser processing device, in order to accurately detect the position of the defect on the processed object and correct the defect, it is required to accurately irradiate the laser light to the necessary target position.

Under the current situation, the defect position on the processed object or the target location of the laser light is found by the human body, that is, the operator visually and analyzing the image of the processed object. In addition, the operator selects the target position of the irradiated laser light, the wavelength of the laser light, the output intensity, the beam diameter or the size, and the projection of the laser light through manual operation through a joystick (joystick) or other operation input device Recipe for shape, focal distance, length of irradiation time, etc., perform laser processing on the processed object.

In factories and other sites, in order to process a large number of processed objects within a certain period of time, multiple laser processing devices are installed. Only in this part, the burden on the operators has increased, and a few more operators have to be deployed.

The desired purpose of the present invention is to promote the automation of part or all of the operations of a plurality of laser processing devices. [Means to solve the problem]

In order to solve the above-mentioned problems, the present invention constitutes a mechanical learning system for a laser processing device. The laser processing device includes: a learning data storage unit that acquires and memorizes defects that should be corrected in the processed object. The target position is irradiated with laser light. Each of the multiple laser processing devices is equipped with the image captured by the visual sensor for shooting the processed object, and at least the defect location of the processed object on the image is suggested Or it is the group of additional information of the target position that should be irradiated with the laser light, that is, the learning data; and the learning unit, which uses the learning data group memorized by the aforementioned learning data storage unit to integrate the visual sensor provided in the laser processing device The image obtained by shooting the object to be processed is used as input, and a learned model is generated that takes at least the defect position or the target position in the image as the output.

The additional information of the learning data memorized in the learning data storage unit includes: an identifier for identifying the content of the laser processing performed by the laser processing device on the object to be processed, or information suggesting the content of the processing. Next, the output of the learned model generated by the aforementioned learning unit includes information specifying the identifier of the content of the laser processing performed by the laser processing device on the object to be processed or the content of the hint processing.

As a specific one of the states, for example: the control controller attached to the laser processing device and the servo computer connected through a telecommunication line to be able to communicate, equipped with the aforementioned learning data storage unit and the aforementioned learning unit; the aforementioned servo computer from the plural The laser processing device receives and collects the aforementioned learning data through the aforementioned control controller, generates the aforementioned learned model, and when controlling each laser processing device, sends the aforementioned learned model generated by the aforementioned servo computer to the attached The aforementioned control controller in each laser processing device.

As another aspect, for example, there is a control controller attached to the laser processing device and a servo computer connected through a telecommunication line to be able to communicate, with the aforementioned learning data storage unit and the aforementioned learning unit; the aforementioned servo computer from a plurality of seats The laser processing device receives and collects the learning data through the control controller to generate the learned model, and when controlling each laser processing device, the servo computer receives the information of each laser processing device through the control controller The image captured by the visual sensor responds to the laser processing device attached to the laser processing device based on the output of the result of giving the image as an input to the aforementioned learning model, and responds to the generated image The controller is used for control.

As further other aspects, for example, there are: a control controller attached to the laser processing device, including the aforementioned learning data storage unit and the aforementioned learning unit; a control controller attached to the laser processing device, and a telecom The line is connected to a servo computer that can communicate. It receives and collects the aforementioned learning data from a plurality of laser processing devices through the aforementioned control controller, and sends the learning data to the control controller attached to each laser processing device Do allotment.

The mechanical learning method for the laser processing device of the present invention has the following steps: acquiring and memorizing each of the multiple laser processing devices that perform the processing of irradiating laser light at the target position where the defect generated in the processed object should be corrected A group of additional information that is provided with the image taken by the visual sensor used to photograph the object to be processed, and the additional information that suggests at least the defect position of the object to be processed on the image or the target position that should be irradiated with laser light, that is, the learning material The steps; and using the aforementioned learning data group, the image obtained by the visual sensor of the laser processing device taken by the processed object is used as input, and at least the defect position or the target position in the image is generated as the output learning Complete the steps of the model. [Effects of the invention]

According to the present invention, it is possible to promote the automation of part or all of the operations of a plurality of laser processing devices.

An embodiment of the present invention will be described with reference to the drawings. At the beginning, the laser processing device 1 to be controlled by the system of this embodiment will be described. The laser processing device 1 in this embodiment is used in a liquid crystal display module, a plasma display module, an organic electroluminescent display module, an inorganic EL display module, a micro LED display module, a transparent conductive film substrate or It is a laser repairing device that irradiates the defect of the color filter and other 0 with laser light L to correct the defect of the workpiece W.

Fig. 1 and Fig. 2 show the configuration of a single laser processing device 1. The main components of the laser processing device 1 are as follows: an irradiation unit 2 that irradiates the laser light L toward the target position on the processed object W, that is, a defect of the liquid crystal display module, etc., moves the irradiation unit 2 to the light of the laser light L The XY stage 3, 7 of the two-dimensional directions of the X-axis and Y-axis, which are orthogonal or slightly orthogonal, move the irradiation unit 2 on the Z-axis parallel to the optical axis of the laser light L or slightly parallel to the Z-axis direction The adjustment mechanisms 3 and 6 and the displacement meter 4 which measures the distance from the irradiation unit 2 to the object W to be processed.

The holder 9 supports the irradiation unit 2, the XY stages 3, 7, the Z-axis adjustment mechanisms 3, 6 and the displacement meter 4, and the object W to be processed. The bracket 9 is grounded to the floor through the anti-vibration member 91. The anti-vibration member 91 is, for example, a passive suspension such as anti-vibration (vibration-damping) rubber or an air spring, and its role is to suppress transmission of vibration with a frequency higher than a specific value (for example, 5 Hz) from the floor to the bracket 9.

The to-be-processed object W is fixed by the holder 9 by other suitable means, such as clamping and suction. In the process of irradiating the treatment object W with laser light L, the treatment object W is a fixed object.

The XY stages 3 and 7 in the laser processing device 1 are a combination of the first XY stage 3 and the second XY stage 7. The second XY platform 7 is set up on the bracket 9 to support the irradiation unit 2, the first XY platform 3, the Z-axis adjustment mechanism 3, 6 and the displacement meter 4, which can move these two dimensions in the X-axis direction and the Y-axis direction. direction. For example, the second XY stage 7 has the following components: a pair of Y-axis rails 71 erected on both sides of the bracket 9 and extending in the Y-axis direction, and a pair of Y-axis type motors that run along each Y-axis rail 71 The trolley 72 and the X-axis rail 73 extending along the X-axis direction supported by the Y-axis motor trolley 72 on both sides, and the X-axis motor trolley 74 running along the X-axis rail 73 respectively. Next, the X-axis type motor trolley 74 supports the irradiation unit 2, the Z-axis adjustment mechanisms 3 and 6, and the displacement meter 4. The irradiation unit 2 and the displacement meter 4 face the object W from directly above the object W in the Z-axis direction. In summary, the second XY stage 7 moves the irradiation unit 2, the Z-axis adjustment mechanisms 3, 6 and the displacement meter 4 in the X-axis direction and the Y-axis direction relative to the holder 9 and the workpiece W.

A linear encoder is attached to the set of the Y-axis rail 71 and the Y-axis motor cart 72, and the position coordinate of the Y-axis motor cart 72 in the Y-axis direction is detected via the linear encoder. A linear encoder is also attached to the set of the X-axis rail 73 and the X-axis motor carriage 74, and the current position coordinate of the X-axis motor carriage 74 in the X-axis direction is detected through the linear encoder. As a result, through the two-line encoder, the current XY position coordinates of the X-axis motor trolley 74 supporting the irradiation unit 2 are detected.

The Z-axis adjustment mechanisms 3 and 6 in the laser processing device 1 are a combination of the first Z-axis adjustment mechanism 3 and the second Z-axis adjustment mechanism 6. The second Z-axis adjustment mechanism 6 is interposed between the frame 5 that integrates the irradiation unit 2, the first XY stage 3, and the displacement gauge 4, and the above-mentioned X-axis type motor trolley 74, so that the frame 5 can be opposed to each other. The X-axis type motor trolley 74 moves in the Z-axis direction. The height position along the Z-axis direction of the X-axis type motor trolley 74 is unchanged. The second Z-axis adjustment mechanism 6 is, for example, a known screw feeding mechanism including a ball screw. One of the X-axis motor trolley 74 and the frame 5 bears the screw shaft, and is fixed and screwed to the other side of the X-axis motor trolley 74 The screw cap of the screw shaft rotates and drives the screw shaft via a servo motor, a stepping motor, etc., thereby causing the nut to advance and retreat along the screw shaft, causing the frame body 5 to move up and down in the Z-axis direction. The second Z-axis adjustment mechanism 6 moves the irradiation unit 2 and the displacement meter 4 in the Z-axis direction relative to the holder 9 and the workpiece W.

A linear encoder is attached to the set of the X-axis motor cart 74 and the frame body 5, and the current position coordinate of the frame body 5 in the Z-axis direction is detected through the linear encoder.

As shown in Figure 2, the irradiation unit 2 housed in the housing 5 includes: an optical system for observing the target position on the processed object W and its surrounding area, and an optical system for irradiating the target position on the processed object W with thunder The optical system of the light L. The former optical system at least includes: an epi-illumination light source 21, a beam splitter (or a half lens) 22, a bidirectional beam splitter 23, an objective lens 24, an imaging lens 251, and a vision sensor 25. The epi-light system supplied from the epi-illumination light source 21 is reflected via the beam splitter 22 and faces the direction of the optical axis of the objective lens 24 facing the object W to be processed. After the falling light passes through the dichroic mirror 23, it irradiates the target position on the processed object W and its peripheral range through the objective lens 24. The luminous flux reflected by the object to be processed is incident on the objective lens 24, and incident on the imaging lens 251 through the dichroic beam splitter 23 and the beam splitter 22, and the image is imaged on the visual sensor 25, which is CCD (Charge-Coupled Device) or CMOS (Complementary Metal-Oxide- Semiconductor) and other solid-state imaging elements. In this way, through the visual sensor 25, the target position on the processed object W and its surrounding area can be photographed and an image can be obtained.

The objective lens 24 has a plurality of different magnifications. These pairs of objective lenses 24 are mounted on a motorized lens exchange rotating base 241, and any one of the objective lenses 24 can be selected and arranged on the optical axis through the rotating lens exchange rotating base 241. That is, the applied magnification of the objective lens 24 can be changed.

The latter optical system includes at least: the light source of the laser light L, that is, an oscillator 26, an attenuator 27, a polarizing plate 28, a light flux expander 29, a variable slit 20, a dichroic mirror 23, and an objective lens 24. The dichroic mirror 23 and the objective lens 24 are common to the lenses of the former optical system. The laser light L supplied from the laser oscillator 26 is attenuated by the attenuator 27, polarized by the polarizing plate 28, and shaped by the luminous flux expander 29 and the variable slit 20. The variable slit 20 can change the shape of the laser light flux passing through these. Furthermore, the laser light L is reflected by the dichroic mirror 23 and faces the direction of the optical axis of the objective lens 24 opposite to the object W to be processed. The laser light L is focused on the target position on the object W by the objective lens 24. In this way, the laser light L can be irradiated to the target position on the object W to be processed.

However, both the optical system of the former and the optical system of the latter may include other optical elements not shown in FIG. 2, such as optical fibers or lenses, lenses, lenses, switches, etc.

The displacement meter 4 is used to measure the distance in the Z-axis direction from the objective lens 24 of the irradiation unit 2 to the target position on the object W to be processed. For example, the displacement meter 4 irradiates the laser light R toward the processed object W and receives the laser light R reflected by the processed object W. Based on this, the distance from the displacement meter 4 to the processed object W is accurately measured. The laser displacement meter. The displacement meter 4 is housed in the frame 5 or attached outside the frame 5. Through the function of the second Z-axis adjustment mechanism 6 described above, the displacement meter 4 is integrated with the housing 5, the irradiation unit 2, and the first XY stage 3, and moves up and down in the Z-axis direction.

The first XY stage 3 is housed in the frame body 5, and supports at least the objective lens 24 and the lens-changing rotating seat 241 in the irradiation unit 2 so that it can move in the two-dimensional direction of the X-axis direction and the Y-axis direction. The first XY stage 3 is, for example, a known piezoelectric stage that precisely moves the stage supporting the objective lens 24 and the lens exchange base 241 in the X-axis direction and the Y-axis direction via a piezoelectric motor (ultrasonic motor). The first XY stage 3 displaces the objective lens 24 of the irradiation unit 2 in the X-axis direction and the Y-axis direction with respect to the frame 5 and further with respect to the holder 9 and the object W to be processed.

The first Z-axis adjustment mechanism 3 is also housed in the frame 5, and supports at least the objective lens 24 and the lens-changing rotating seat 241 in the irradiation unit 2 so that it can move in the Z-axis direction. The first Z-axis adjustment mechanism 3 is, for example, a known piezoelectric stage that precisely moves the stage supporting the objective lens 24 and the lens exchange rotating seat 241 in the Z-axis direction via a piezoelectric motor. The first Z-axis adjustment mechanism 3 displaces the objective lens 24 of the irradiation unit 2 in the Z-axis direction relative to the frame 5 and further relative to the holder 9 and the object W to be processed.

Furthermore, in this embodiment, the two functions of the first XY stage 3 and the first Z-axis adjustment mechanism 3 are realized through a single piezoelectric motor XYZ stage. The piezoelectric XYZ stage 3 is capable of moving the stage supporting the objective lens 24 and the lens-changing rotation base 241 in the three-dimensional directions of the X-axis direction, the Y-axis direction and the Z-axis direction.

Elements other than the objective lens 24 of the irradiation unit 2 or the displacement meter 4 do not necessarily need to be mounted on the piezoelectric XYZ stage 3. It is possible to reduce the stowage of the piezoelectric XYZ stage 3, and to improve the accuracy and responsiveness of the position control of the objective lens 24 by the piezoelectric XYZ stage 3 as much as possible. However, it does not exclude that at least a part of the elements other than the objective lens 24 of the irradiation unit 2 is mounted on the piezoelectric XYZ stage 3 and moved together with the objective lens 24, nor is it excluded that the position shift meter 4 is mounted on the piezoelectric When the XYZ stage 3 moves together with the objective lens 24.

An optical linear encoder is attached to the piezoelectric XYZ stage 3 as the first XY stage and the first Z-axis adjustment mechanism, and the position coordinates of the stage in the X-axis direction, Y-axis direction and Z-axis direction are detected through the linear encoder.

In this embodiment, the width dimension along the X-axis direction and the Y-axis direction of the object W to be processed is assumed to exceed 1 m. In order to be able to irradiate the laser light L to substantially the entire area of the object W of such a large size, the stroke of the second XY stage 7, that is, the movable range in the two-dimensional direction is 1 m or more. That is, the movable range of the Y-axis direction of the Y-axis type motor bogie 72 is 1 m or more, and the movable range of the X-axis direction of the X-axis type motor bogie 74 is also 1 m or more.

Conversely, the stroke of the first XY stage 3, that is, the movable range in the X-axis direction and the Y-axis direction is much smaller than that of the second XY stage 7, and supports the X-axis and Y-axis directions of the stage of the objective lens 24. The movable range is below 100μm, at most about 40μm to 50μm. Even so, the minimum displacement of the first XY stage 3 is still very small compared to the second XY stage 7. Compared with the second XY stage 7, the first XY stage 3 can finely and precisely adjust the position of the objective lens 24 in the X-axis direction and the Y-axis direction. In addition, the resolution of the position coordinate detection caused by the optical linear encoder attached to the first XY stage 3 is higher than that of the position coordinate detection caused by the linear encoder attached to the second XY stage 7 The resolution is even higher.

Moreover, the stroke of the first Z-axis adjustment mechanism 3, that is, the movable range in the Z-axis direction is much smaller than that of the second Z-axis adjustment mechanism 6, and the movable range in the Z-axis direction of the stage supporting the objective lens 24 is 10 μm to About 30μm. Even so, the minimum displacement of the first Z-axis adjustment mechanism 3 is still very small compared to the second Z-axis adjustment mechanism 6. The first Z-axis adjustment mechanism 3 can finely and precisely adjust the position of the objective lens 24 in the X-axis direction and the Y-axis direction than the second Z-axis adjustment mechanism 6. In addition, the resolution of the position coordinate detection caused by the optical linear encoder attached to the first Z-axis adjustment mechanism 3 is higher than that of the linear encoder attached to the second Z-axis adjustment mechanism 6 The resolution of the coordinate detection is even higher.

The control controller 8 that is attached to the laser processing device 1 and controls the control of the laser processing device 1 is, for example, a general-purpose personal computer or a work station as the main body. As shown in FIG. 3, the control controller 8 is a hardware including a CPU (Central Processing Unit) 81, a main memory 82, an auxiliary memory device 83, a video codec 84, a display 85, a communication interface 86, an operation input device 87, etc. Physical resources, and act in conjunction with these hardware players.

The auxiliary memory device 83 is a flash memory, a hard disk device, an optical disk device, etc. The main components of the video codec 84 are as follows: according to the drawing instructions received from the CPU 83, the GPU (Graphics Processing Unit) that generates the picture to be displayed, and sends the picture signal to the display 85; and temporarily stores the picture or image Video memory for data, etc. The video codec 84 may not be hardware, but may be installed as software. The communication interface 86 is a device used by the control controller 8 for information communication with external devices 0 and 1, including a wired connection interface or a wireless transceiver. The operation input device 87 is a keyboard, a pointing device for pressing a button, a joystick, a mouse, or a touch panel (which is also overlapped with the display 85) operated by the operator with a finger, or other devices.

In the control controller 8, the program to be executed via the CPU 81 is stored in the auxiliary memory device 83, and when the program is executed, the program is read from the auxiliary memory device 83 to the main memory 82 and read by the CPU 81. The control controller 8 activates the hardware resources in accordance with the program to control the laser processing device 1.

FIG. 4 shows an example of the processing sequence performed by the control controller 8 in the laser processing by the laser processing device 1. First, the controller 8 for control prepares to irradiate the laser light L from the irradiation unit 2 of the laser processing device 1 toward the target position on the object W to be processed, and to give a control signal to the second XY stage 7 to pass the light on the objective lens 24. The axis, that is, the irradiation position of the laser light L that irradiates the object lens 24 to the processed object W coincides with or is close to the target position on the processed object W. To this end, the second XY stage 7 is driven to include the irradiation unit 2 The frame body 5 moves in the X-axis direction and/or the Y-axis direction (step S1).

Next, the control controller 8 gives a control signal to the first XY stage 3, and the irradiation position of the laser light L irradiated to the object W by the objective lens 24 precisely coincides with the target position on the object W. , The first XY stage 3 is driven, and the position of the objective lens 24 of the irradiation unit 2 in the X-axis direction and/or the Y-axis direction is fine-tuned (step S2). At this time, through the visual sensor 25 of the irradiation unit 2, the position that intersects the optical axis of the objective lens 24 in the object W and the surrounding area are captured, the obtained image is analyzed, and the surface of the object W is detected. Correctly align the optical axis to its target position, and the displacement caused by the first XY stage 3 can be slightly corrected. However, the imaging of the processed object W by the visual sensor 25 is preferably performed after the focus of the objective lens 24 is focused on the processed object W.

Almost at the same time as the above step S2, the control controller 8 gives control signals to the Z-axis adjustment mechanisms 3 and 6, to focus on the objective lens 24 and then irradiate the laser light to the object W through the pair of objective lenses 24 The focal point of L is precisely focused on the target position on the object W. For this reason, the Z-axis adjustment mechanisms 3 and 6 are driven to fine-adjust the frame 5 including the irradiation unit 2 and/or the Z-axis of the objective lens 24 The position of the direction (step S3). In this step S3, the function of the displacement meter 4 is not used. In step S3, for example, through the visual sensor 25 of the irradiation unit 2, the position that intersects the optical axis of the objective lens 24 in the object W and the surrounding area are repeatedly captured, and the obtained images are analyzed and successively The contrast (difference between light and dark) is determined, and the frame 5 and/or the objective lens 24 are moved up and down until the contrast of the photographed image becomes the maximum or close to the maximum height position.

The reason why the displacement meter 4 is not used in step S3 is that, as shown in FIG. 2, in order to perform laser processing on the processed object W, the irradiation position of the laser light L irradiated by the objective lens 24 and the displacement meter 4 The irradiated positions of the laser light R irradiated at a distance up to the workpiece W are not coincident with each other but deviate. The larger the magnification of the objective lens 24, the larger the size of the objective lens 24 itself. In order to avoid interference with the objective lens 24, the displacement meter 4 has to be arranged far away from the objective lens 24. In short, it is difficult to match the optical axis of the displacement meter 4 to the optical axis of the objective lens 24. The large workpiece W may be deformed integrally or partially. Considering that the degree of flexural deformation is not constant, it is impossible to directly regard the distance measured by the displacement meter 4 as the target position on the object W from the objective lens 24. Distance. That is, the position of the distance measured by the displacement meter 4 is different from the position where the laser light L emitted from the objective lens 24 hits, and it is caused by the deflection of the irradiated object along the Z axis of the two positions. The height of the direction is not the same possibility, and the height difference between the two positions cannot be known in advance. Therefore, in step S3, the displacement meter 4 cannot be used, but the contrast AF (Auto Forcus) method or the like is used to focus the focus of the laser light L to the irradiation position on the object W to be processed.

After the above steps S1 to S3, the objective lens 24 of the irradiating unit 2 reaches the basic position necessary for the processing of irradiating the laser light L to the target position on the object W to be processed. Thereafter, the control controller 8 starts feedback control for maintaining the objective lens 24 at the basic position (steps S4 and S5). The position feedback control is performed to prevent the laser light L from being irradiated to the processed object W when the optical axis or focus of the laser light L deviates from the target position on the processed object W due to interference. A typical disturbance is the vibration transmitted from the floor of the factory etc. where the laser processing device 1 is installed to the bracket 9, the workpiece 0, the XY stage 7, the Z-axis adjustment mechanism 6, and the irradiation unit 2. The vibration with a high frequency (over 5 Hz) to a certain extent is blocked or sufficiently attenuated by the anti-vibration member 91 between the floor and the bracket 9. However, the low frequency (less than 5 Hz, about 2 Hz to 3 Hz) vibration is not necessarily blocked by the anti-vibration member 91, but can be transmitted from the floor to the bracket 9, the workpiece 0, the XY stage 7, and the Z-axis adjustment mechanism 6 and irradiation unit 2.

In the position feedback control step S4 in the X-axis direction and the Y-axis direction, the control controller 8 sets the position of the irradiation unit 2 to the basic position, and then repeatedly shoots and is photographed through the visual sensor 25 of the irradiation unit 2 The position where the optical axis of the objective lens 24 intersects and its peripheral range in the processed object W are analyzed, and the obtained image is analyzed, and the position of the irradiation unit 2 appears successively from the basic position, that is, from the irradiation position of the laser light L. How much deviation does the position at the target position deviate from in the X-axis direction and the Y-axis direction. Then, the control signal is given to the first XY stage 3, and the first XY stage 3 is driven to reduce the deviation, and the position of the objective lens 24 is corrected toward the basic position. With this, the optical axis of the objective lens 24, in other words, the irradiation position of the laser light L, can be continuously maintained at or near the target position on the object W to be processed.

In step S5 of the position feedback control in the Z-axis direction, when the control controller 8 positions the irradiation unit 2 to the basic position, the displacement meter 4 measures the distance to the object W to be processed, and sets the distance as feedback The target distance of the control. In addition, thereafter, the displacement meter 4 repeatedly measures the separation distance to the object W to be processed, and successively obtains the deviation between the measured current separation distance and the target distance. This deviation indicates how much the current position of the irradiation unit 2 deviates in the Z-axis direction from the basic position, that is, the position when the focal point of the laser light L coincides with the target position. Next, the control signal is given to the first Z-axis adjustment mechanism 3, and the first Z-axis adjustment mechanism 3 is driven to correct the position of the objective lens 24 toward the basic position in the direction of reducing the deviation. Through this, the focus of the laser light L can be continuously maintained at or near the target position on the object W to be processed.

Through the position feedback control, it is possible to quickly correct the deviation caused by the low-frequency vibration transmitted from the floor to the bracket 9, the workpiece 0, the XY stage 7, the Z-axis adjustment mechanism 6, and the irradiation unit 2. The piezoelectric motor stage 3, which is the first XY stage and the first Z-axis adjustment mechanism, has a sufficient response speed in order to correct the deviation caused by the low-frequency vibration.

In addition, the movable range of the piezoelectric motor stage 3 supporting the objective lens 24 is 100 μm or less, which is very small. In addition, the laser light L supplied from the laser oscillator 26 and incident on the objective lens 24 is collimated parallel light. Therefore, when components other than the objective lens 24 of the irradiation unit 2 are not mounted on the piezoelectric motor stage 3, when the objective lens 24 is displaced in the X-axis direction or the Y-axis direction through the piezoelectric motor stage 3, only the irradiated object is displaced The position where the visual sensor 25 takes the image on 0 and the position where the laser light L is irradiated are the same amount. Even if the objective lens 24 is displaced in the Z-axis direction through the piezoelectric motor stage 3, there is no problem.

In this way, the control controller 8 performs the placement feedback control on the one hand, and irradiates the laser light L to the target position on the object W on the other hand (step S6).

When laser processing the object W, it is necessary to determine the wavelength, output intensity, beam diameter or size of the laser light L used to irradiate the target position of the laser light L, and the projection shape and focus of the laser light L The formula of distance, length of irradiation time, etc. In this embodiment, the learned model generated by machine learning is used to analyze and photograph the image of the liquid crystal display panel or the like of the processed object W to determine the position of the defect in the processed object W, and then determine the application The XY coordinates of the target position where the laser light L is irradiated. Furthermore, the wavelength, output intensity, beam diameter or size of the laser light L in the laser processing, the projection shape of the laser light L, the focal length, etc. are also set. The focal distance is equivalent to the Z coordinate of the target position.

As shown in FIG. 5, the machine learning system of this embodiment is connected to a control controller 8 attached to a plurality of laser processing devices 1 installed in a factory, etc., through wired LAN (Local Area Network) or wireless LAN, WAN (Wide Area Network), MAN (Mertopolitan Area Network), portable telephone network, Internet and other telecommunication lines are connected to the servo computer and can be constructed to communicate with each other. Moreover, each of the plurality of laser processing devices 1 is attached with a control controller 8, and each control controller 8 is not limited to be in charge of the control of the corresponding laser processing device 1. One control controller 8 can also control the laser processing apparatus 1 of two or more. Furthermore, the specific control controller 8 may also serve as the task of the servo computer 0.

The servo computer 0 is composed of, for example, a general-purpose personal computer or a work station as its main body. As shown in Figure 6, the server computer 0 has hardware resources such as CPU01, main memory 02, auxiliary memory device 03, video codec 04, display 05, communication interface 06, operation input device 07, etc. These are joint actions.

The auxiliary memory device 03 is a flash memory, a hard disk device, an optical disk device, etc. The main components of the video codec 04 are as follows: according to the drawing instructions received from the CPU03, the GPU that generates the screen that should be displayed, and sends the screen signal to the display 05; and the video memory that temporarily stores the screen or image data Wait. The video codec 04 can also be installed as software instead of hardware. The communication interface 06 is a device used by the server computer 0 for information communication with an external device 8, including a wired connection interface or a wireless transceiver. The operation input device 07 is a pointing device operated by an operator with a finger, a keyboard, pressing a button, a joystick, a mouse or a touch panel (also overlapped with the display 05), or other devices.

In the servo computer 0, the program to be executed via the CPU01 is stored in the auxiliary memory device 03, and when the program is executed, it is read from the auxiliary memory device 03 to the main memory 02 and interpreted by the CPU01.

In one of the specific states shown in FIG. 7, the servo computer 0 operates the hardware resources according to the program, and functions as the learning data storage unit 101 and the learning unit 102 described below.

The learning data storage unit 101 uses the main memory 02 or the auxiliary memory of the learning data generated by the control controller 8 in charge of the control of the laser processing device 1 from each of the plurality of laser processing devices 1 The required memory area of the device 03 to be memorized. The learning material is a combination of: the visual sensor 25 of each laser processing device 1 captures the image obtained by the processed object W; and the addition of a recipe that defines the content of the laser processing to be performed on the same processed object W News.

As shown in Figure 8, the learned model to be generated is when the image of the processed object W that is the target of laser processing is input to the input layer, the input passes through the intermediate layer (hidden layer) to the output layer, and becomes a definition pair. A neural network (or, deep learning artificial intelligence) output with the formula information of the content of the laser processing of the processed object W. The photographic image of the processed object W in the learning material corresponds to the input of this type of neural network. Then, the additional information in the learning materials corresponds to the output of the same kind of neural network, which conforms to the training materials in the learning of the accompanying teacher.

Therefore, the additional information includes at least: the coordinates (or the range of coordinates) indicating the defect position of the processed object W in the photographic image of the processed object W, and/or for correcting the defect of the processed object W Instead, the coordinates (or the range of the coordinates) of the target position where the laser light L is irradiated with the object W to be processed.

Here, the defect position of the workpiece W and the target position of the laser beam L are not always the same, but they may be different. Even if the defect location is the same, the target location will change because of the type or degree of the defect. For example, when any defect occurs in a certain pixel of the liquid crystal display panel W, laser light L is irradiated to the short circuit of the wiring connected to the pixel to cut off the short circuit, or laser light is irradiated to the broken wire. L is used to weld the short-circuit, or the foreign matter mixed in the manufacturing process is irradiated with laser light L to remove the foreign matter, thereby eliminating the defect, or applying laser light L to the defective pixel itself. Let the pixel light up so that it becomes inconspicuous. If the defect position and the target position on the processed object W are the same, or if the target position can be specified without distinction from the defect position, there is no need to include the target position in the additional information (and the output of the learned model). However, this is not necessarily the case, it is necessary to include the target location in the additional information, or to include any other information for the specific target location in the additional information.

Furthermore, the additional information includes some or all of the following items, etc.: ・In addition to the information indicating the type and/or degree of the defect, other information that implies the relative positional relationship between the defect location and the target location ・Wavelength of the laser light L used ・The output of the laser light L used ・The harness diameter or size of the laser light L used ・The shape of the harness of the laser light L used ・The focal length of the laser light L used ・The length of time the laser light L is irradiated to the object W to be processed.

Or, the additional information is a group of each uplifted item, that is, any identifier among the candidates including a specified plural number of recipes. In formula α and formula β, at least a part of the relative positional relationship between the defect position and the target position, the wavelength, output, beam diameter or size of the laser light L, the projection shape of the beam, the focal length, the irradiation time, etc. are different . The identification sub-system indicates whether the target position on the processed object W mapped on the image that becomes a part of the learning material is to be subjected to the laser treatment caused by the formula α or the laser treatment caused by the formula β.

The additional information contained in the learning material is input to the control controller 8 by an operator who has visually viewed the photographic image of the processed object W contained in the same learning material through the display 85 by operating the operation input device 87. For example, the operator designates the defect position (range) and/or the target position (range) shown in the image, or selects the necessary mine from a plurality of candidates in order to correct the defect of the processed object W. The formula of the laser treatment, or specify the wavelength of the laser light L used, and output other parameters.

The control controller 8 that has received the input of the additional information related to the processed object W sends the additional information and the photographic image of the processed object W paired with the additional information to the servo computer 0 as learning data. The servo computer 0 receives learning data generated from the control controller 8 of the plurality of laser processing devices 1, collects learning data related to a plurality of objects W to be processed, and accumulates the learning data storage unit 101.

The learning part 102 uses the learning material group memorized by the learning material memory part 101 to generate a learning completion model. That is, the learning materials stored in the learning material storage unit 101 are sequentially read out, a photographic image of the processed object W contained in a certain learning material is input as input, and the defect position on the processed object W reflected in the image is taken as input (The range) and/or the target position (the range) or other additional information is used as output. In order to make such output for such input, deep learning that makes neural network-like changes is performed. The content of the additional information output by the similar neural network is the same as the content contained in the learning materials used in machine learning. By the way, in the machine learning, the servo computer 0 may use the calculation ability of the GPU.

The servo computer 0 of the learned model generated through the above-mentioned mechanical learning sends the data of the learned model to the control controller 8 attached to each of the plurality of laser processing devices 1 for distribution. Each control controller 8 receives the data of the learned model generated from the servo computer 0, and stores it in the main memory 82 or the required memory area of the auxiliary memory device 83.

The learned model system can be used in the future to perform laser processing on the object W via the laser processing device 1. In other words, the control controller 8 acquires an image of the object W that has been moved into the corresponding laser processing device 1 taken through the visual sensor 25, and provides the image as an input to the neural network of the learned model By this, the position (or the range) of the defect existing on the processed object W and/or the position (or the range) of the target to be irradiated with the laser light L can be output as the output. Furthermore, the content contained in the learning materials used in machine learning can be the same type of additional information, that is, the information indicating the type and/or degree of the defect, or the wavelength of the laser light L used, The output intensity, beam shape, focal distance, etc., or the information of the identifier that identifies the candidate for the recipe for which the laser processing of the object W should be executed is output as output. Then, the control controller 8 irradiates the object W with laser light L based on the recipe determined using the learned model.

In the present embodiment, a mechanical learning system for the laser processing device 1 is constituted. The laser processing device includes: a learning data storage unit 101 that acquires and stores the defects that should be corrected in the processed object W. The image captured by the visual sensor 25 for photographing the processed object W provided in each of the plurality of laser processing devices 1 for processing the laser light L irradiated at the target position, and hints at least the processed object on the image The defect position of W or the set of additional information of the target position to be irradiated with laser light, that is, the learning data; and the learning unit 102, which uses the learning data group memorized by the aforementioned learning data storage unit 101 to integrate the laser processing device 1 The included visual sensor 25 takes an image of the processed object W as an input, and generates a learned model whose output is at least a defect position or a target position in the image. According to this embodiment, when each of the plurality of laser processing devices 1 performs laser processing on the processed object W, using the learned model, it is possible to automatically detect the position of a defect in the target processed object W without relying on manpower. , Or the laser light L should be irradiated to the target position of the object to be processed W.

The additional information of the learning data stored in the aforementioned learning data storage unit 101 includes: identifying the content of the laser processing that the laser processing device 1 should perform on the object to be processed W (except for indicating the type of defect and/or Other information indicating the relative positional relationship between the defect position and the target position, or the identifier (specified) of the wavelength, output, wire diameter or size of the laser light L used, the projection shape of the wire harness, the focal length, etc.) Recipe identifier) or information that implies the content of the processing itself; the output of the learned model generated by the aforementioned learning unit 102 includes: the laser processing that the designated laser processing device 1 should perform on the object to be processed W The identifier of the content or the information that implies the content of the processing. As a result, when each of the plurality of laser processing devices 1 performs laser processing on the processed object W, it is possible to automatically determine the use of the learned model without relying on manpower to perform laser processing on the target processed object W Part or all of the content.

The machine learning system of this embodiment can automate part or all of the operations of a plurality of laser processing devices 1.

Furthermore, in the system of this embodiment, the control controller 8 attached to the laser processing device 1 and the servo computer 0 that can communicate via a telecommunication line are provided with the aforementioned learning data storage unit 101 and the aforementioned learning unit 102; The servo computer 0 receives and collects the learning data from the plurality of laser processing devices 1 through the control controller 8 to generate the learned model, and when controlling each laser processing device 1, the servo computer 0 is generated The learned model is sent to the control controller 8 attached to each laser processing device 1. With this, the results of the mechanical learning can be distributed to the control controller 8 of the plurality of laser processing devices 1, and the results can be utilized for the control of the plurality of laser processing devices 1. This greatly contributes to efficiency improvement in factories and the like that operate the plural-based laser processing device 1.

Furthermore, the present invention is not limited to the embodiments detailed above. In the above embodiment, the learned model generated by the servo computer 0 is sent to the plurality of control controllers 8 attached to the plurality of laser processing devices 1, and each control controller 8 uses the model to control the laser processing The laser processing by the device 1 is, in particular, the detection of the defect position on the processed object W or the determination processing of other recipes for irradiating the target position of the laser light L.

However, the main body of the decision process that executes the recipe using the learned model is not limited to the control controller 8. The servo computer 0 keeps the generated data of the learned model in the main memory 02 or the required memory area of the auxiliary memory device 03, which does not hinder the determination of the main body of the formula decision process.

More specifically, the laser processing of the object W caused by each of the plurality of laser processing devices 1 is performed. When each laser processing device 1 is controlled, the visual sensor 25 of each laser processing device 1 The image of the object to be processed W that captured the subject is sent from the control controller 8 to the servo computer 0, and the servo computer 0 receives the image.

Then, in the servo computer 0, the image is given as an input to the neural network of the learned model, thereby, the position of the defect existing on the processed object W and/or the target that should be irradiated with the laser light L As output. Furthermore, the content contained in the learning materials used in machine learning can be the same type of additional information, that is, the information indicating the type and/or degree of the defect, or the wavelength of the laser light L used, The output intensity, beam shape, focal distance, etc., or the information of the identifier that identifies the candidate for the recipe for which the laser processing of the object W should be executed is output as output. The servo computer 0 responds to the output information to the control controller 8 for generating a photographic image of the processed object W. The control controller 8 receives the output information generated from the servo computer 0, and irradiates the processed object W with laser light L based on the recipe determined based on the information.

Furthermore, in the above-mentioned embodiment, the servo computer 0 includes the learning data storage unit 101 and the learning unit 102, and collects the learning data generated from the control controller 8 attached to each laser processing device 1 to the servo computer 0. The machine learning that generates the learned model is executed on the servo computer 0.

On the other hand, as shown in FIG. 9, it is also considered that each control controller 8 has a learning data storage unit 101 and a learning unit 102, that is, it is not the servo computer 0 but each control controller 8 executes the generation and learning. The mechanical learning state of the model.

In this case, the servo computer 0 receives and collects learning data from the plurality of laser processing devices 1 through the control controller 8, and sends the learning data to each control controller attached to each laser processing device 1 8 Do allotment. Each control controller 8 stores the learning data received from other control controllers 8 through the servo computer 0 in the main memory 82 as the learning data storage unit 101 or the required memory area of the auxiliary memory device 83 . The important point is that a group of learning materials is shared among the plurality of control controllers 8 that control each of the plurality of laser processing devices 1. The servo computer 0 relays the learning data generated by each control controller 8 to other control controllers 8. Here, the specific control controller 8 also has the task of the servo computer 0.

In this way, the learning unit 102 of each control controller 8 uses the learning material group memorized by the learning material storage unit 101 to generate a learned model. In this machine learning, each control controller 8 utilizes the calculation capability of the GPU. Each control controller 8 stores the generated data of the learned model in a required memory area of the main memory 82 or the auxiliary memory device 83.

When the laser processing of the object W by each laser processing device 1 is to be performed and the laser processing device 1 is to be controlled, the control controller 8 acquires the visual sensor 25 of the laser processing device 1 The image of the processed object W is taken. Next, in the control controller 8, the image is given as an input to the quasi-neural network of the learned model, whereby the position of the defect existing on the processed object W and/or the laser light L should be irradiated The location of the target, as output. Furthermore, the content contained in the learning materials used in machine learning can be the same type of additional information, that is, the information indicating the type and/or degree of the defect, or the wavelength of the laser light L used, The output intensity, beam shape, focal distance, etc., or the information of the identifier that identifies the candidate for the recipe for which the laser processing of the object W should be executed is output as output. The control controller 8 irradiates the object W with laser light L based on the recipe determined based on the output information.

In addition, the specific configuration of each part, the order of processing, etc., can be variously modified without departing from the scope of the present invention. [Industrial availability]

The present invention can be applied to the control of a laser processing device that irradiates a processed object with laser light and performs desired processing.

W: Object to be processed 1: Laser processing device 8: Controller for control 0: Servo computer 101: Learning Materials and Memory Department 102: Learning Department

[Fig. 1] A diagram showing the overall outline of a laser processing device in an embodiment of the present invention. [Fig. 2] A diagram schematically showing the optical system, the first XY stage, and the displacement meter of the irradiation unit of the laser processing device. [Fig. 3] A diagram showing the hardware resources of the control controller of the laser processing device. [Fig. 4] A flowchart showing an example of the procedure of processing executed by the control controller of the laser processing device according to the program. [Fig. 5] A diagram showing the configuration of the machine learning system in the same embodiment. [Figure 6] A diagram showing the hardware resources of the server computer. [Fig. 7] A functional block diagram of the machine learning system in the same embodiment. [Fig. 8] A diagram showing the concept of a learned model generated by the machine learning system of the same embodiment. [Fig. 9] A functional block diagram of a machine learning system related to a modification of the present invention.

0: Servo computer

8: Controller for control

101: Learning Materials and Memory Department

102: Learning Department

Claims (6)

A mechanical learning system for a laser processing device, the laser processing device having: The learning data memory unit, which acquires and memorizes the visual perception of the multiple laser processing devices that perform the processing of irradiating the laser light at the target position where the defect generated in the processed object should be corrected. It is used to photograph the processed object. The image captured by the detector and the group of additional information suggesting at least the defect position of the processed object on the image or the target position that should be irradiated with laser light, that is, the learning material; and The learning unit uses the learning data group memorized by the aforementioned learning data storage unit, and takes the image of the processed object captured by the visual sensor of the laser processing device as input, and generates at least the defect position in the image Or the target position is used as the output of the learned model. Such as the mechanical learning system of claim 1, in which, The additional information of the learning data memorized in the aforementioned learning data storage unit includes: an identifier that identifies the content of the laser processing performed by the laser processing device on the object to be processed or information that implies the content of the processing; The output of the learned model generated by the aforementioned learning unit includes: an identifier specifying the content of the laser processing performed by the laser processing device on the object to be processed, or information suggesting the content of the processing. Such as the mechanical learning system of claim 1, in which, The control controller attached to the laser processing device and the servo computer that can communicate through the telecommunication line are equipped with the aforementioned learning data storage unit and the aforementioned learning unit; The aforementioned servo computer receives and collects the aforementioned learning data from a plurality of laser processing devices through the aforementioned control controller, and generates the aforementioned learned model, and, When controlling each laser processing device, the learned model generated by the servo computer is sent to the control controller attached to each laser processing device. Such as the mechanical learning system of claim 1, in which, The control controller attached to the laser processing device and the servo computer that can communicate through the telecommunication line are equipped with the aforementioned learning data storage unit and the aforementioned learning unit; The aforementioned servo computer receives and collects the aforementioned learning data from a plurality of laser processing devices through the aforementioned control controller, and generates the aforementioned learned model, and, When controlling each laser processing device, the servo computer receives the image captured by the visual sensor of each laser processing device through the controller for control, and gives the result of the learned model based on the image as input. The output control signal for laser processing responds to the control controller attached to the laser processing device that generated the image. Such as the mechanical learning system of claim 1, in which, The control controller attached to the laser processing device is provided with the aforementioned learning material storage unit and the aforementioned learning unit; The control controller attached to the laser processing device and the servo computer that can communicate through telecommunication lines receive and collect the aforementioned learning data from the plurality of laser processing devices through the aforementioned control controller, and transfer the learning data It is sent to the control controller attached to each laser processing device for distribution. A mechanical learning method for a laser processing device has the following steps: Acquire and memorize the images captured by the visual sensor used to photograph the processed object in each of the multiple laser processing devices that implement the processing of irradiating the laser light at the target position where the defect generated in the processed object should be corrected , And a group of additional information that implies at least the defect position of the processed object on the image or the target position of the laser light that should be irradiated, that is, the step of learning materials; and Using the aforementioned learning data group, take the image of the processed object taken by the visual sensor of the laser processing device as input, and generate a learned model with at least the defect position or the target position in the image as the output .

Images