JP7455014B2 - processing equipment - Google Patents

Description

The present invention relates to processing equipment.

2. Description of the Related Art In order to divide a workpiece such as a semiconductor wafer into chip sizes, there is a processing device that processes the workpiece along a planned dividing line. For example, Patent Document 1 discloses a technique of capturing an image of a workpiece and detecting a region to be processed.

Japanese Patent Application Publication No. 2005-166991

When machining is performed along a planned dividing line, an operation called alignment is performed to detect the position of the planned dividing line in advance prior to machining. In order for the processing equipment to automatically perform this alignment, the operator first images the surface of the workpiece, sets a key pattern that is the key to detecting the planned dividing line, and calculates the distance from the key pattern to the planned dividing line. Performs an operation called teaching to register in processing equipment. However, if the operator recognizes a device pattern similar to the planned dividing line as the scheduled dividing line and performs teaching work, or if the operator performs teaching work on a part that is not the planned dividing line due to an incorrect operation, the machining The device could continue to machine the wrong location on the workpiece.

The present invention has been made in view of the above, and an object of the present invention is to provide a processing device that can improve the accuracy of registering dividing lines and prevent processing at incorrect positions.

In order to solve the above-mentioned problems and achieve the objects, the processing apparatus of the present invention is a processing apparatus for processing a workpiece in which a device area is formed on a surface partitioned by a plurality of planned dividing lines, The processing device includes a holding table that holds a workpiece, a processing feed unit that processes and feeds the holding table in the X direction, and a processing unit that forms a processing groove in the workpiece held on the holding table. An indexing unit that indexes and feeds the processing unit in the Y direction intersecting the X direction, an imaging unit that captures an image of the workpiece held on the holding table, and a control unit. Input using a trained program that has learned training data including an OK image in which the planned dividing line is located at the center of the image of the surface of the workpiece, and an NG image in which the scheduled dividing line is not located in the center of the image. A machine learning unit that processes the input image and determines whether the input image is an OK image or an NG image, and a machine learning unit that processes the input image and determines whether the input image is an OK image or an NG image An input section that inputs an image taken by being positioned at the center of the unit to the machine learning section, and a teach section that registers a key pattern formed on the surface of the workpiece, the teach section If the machine learning unit determines that the image input to the machine learning unit is an OK image, the machine learning unit registers the key pattern and the distance of the planned division line from the key pattern, and the machine learning unit determines whether the image is an NG image or not. If it is determined, a warning is issued.

The processing device of the present invention has the effect of improving the accuracy of registering the planned division line and preventing processing at the wrong position.

FIG. 1 is a perspective view schematically showing a configuration example of a processing device according to an embodiment. FIG. 2 is a plan view showing a configuration example of a wafer processed by the processing apparatus according to the embodiment. FIG. 3 is a block diagram schematically showing an example of the functional configuration of the processing device according to the embodiment. FIG. 4 is a diagram for explaining an example of a learned program used by the processing device according to the embodiment. FIG. 5 is a flowchart illustrating an example of a processing procedure regarding registration of image data according to the embodiment. FIG. 6 is a flowchart illustrating an example of a processing procedure regarding determination by the machine learning unit according to the embodiment. FIG. 7 is a flowchart illustrating an example of a processing procedure regarding the teaching unit according to the embodiment. FIG. 8 is a diagram for explaining an example of registration of key patterns and streets according to the embodiment. FIG. 9 is a flowchart illustrating an example of a processing procedure during auto-alignment of the processing apparatus according to the embodiment. FIG. 10 is a perspective view schematically showing a configuration example of a processing device according to a modification of the embodiment. FIG. 11 is a block diagram showing a configuration example of the laser irradiation unit shown in FIG. 10.

DESCRIPTION OF EMBODIMENTS Hereinafter, embodiments according to the present invention will be described in detail with reference to the drawings. The present invention is not limited to the contents described in the following embodiments. Further, the constituent elements described below include those that can be easily assumed by those skilled in the art and those that are substantially the same. Furthermore, the configurations described below can be combined as appropriate. Further, various omissions, substitutions, or changes in the configuration can be made without departing from the gist of the present invention.

In the embodiment described below, an XYZ orthogonal coordinate system is set, and the positional relationship of each part will be explained with reference to this XYZ orthogonal coordinate system. One direction in the horizontal plane is defined as the X-axis direction, a direction perpendicular to the X-axis direction in the horizontal plane is defined as the Y-axis direction, and a direction perpendicular to each of the X-axis direction and the Y-axis direction is defined as the Z-axis direction. The XY plane including the X and Y axes is parallel to the horizontal plane. The Z-axis direction perpendicular to the XY plane is the vertical direction.

[Embodiment]
A processing device according to an embodiment will be described with reference to the drawings. FIG. 1 is a perspective view schematically showing a configuration example of a processing device according to an embodiment. FIG. 2 is a plan view showing a configuration example of a wafer processed by the processing apparatus according to the embodiment. FIG. 3 is a block diagram schematically showing an example of the functional configuration of the processing device according to the embodiment.

As shown in FIG. 1, a processing apparatus 10 (an example of a processing apparatus) according to the embodiment is a cutting apparatus that cuts a wafer 100 along a street (dividing line) 103. The processing apparatus 10 basically includes a chuck table 11 that holds the wafer 100, an imaging unit 12, a processing unit 13, a driving means 14, a Z-axis moving means 15, a touch panel 17, an LED light 18, and a control unit 40. Be prepared.

A wafer (an example of a workpiece) 100 is attached to the surface of an adhesive tape 107 attached to an annular frame 108, as shown in FIG. Further, the wafer 100 is divided into a plurality of regions by a plurality of streets 103 formed in a grid pattern on the front surface 101, and devices 104 having a semiconductor chip structure such as an IC or an LSI are formed in the divided regions. The back surface 102 of the wafer 100 configured in this manner is attached to an adhesive tape 107 attached to an annular frame 108 with the front surface 101 facing upward.

Returning to FIG. 1, the chuck table 11 (an example of a holding table) has a holding surface 11a that attracts and holds the wafer 100, and is rotatably connected to a motor 19. Furthermore, the chuck table 11 is provided so as to be movable in the X-axis direction, which is horizontal to the holding surface 11a, by an X-axis moving means 22 of a well-known configuration including a ball screw 20, a nut, a pulse motor 21, and the like. Thereby, the processing apparatus 10 can move the wafer 100 held on the chuck table 11 in the X-axis direction relative to the imaging unit 12 and the processing unit 13.

The imaging unit 12 is an electron microscope equipped with an image sensor such as a CCD (Charge Coupled Device) or a CMOS (Complementary Metal Oxide Semiconductor). The imaging unit 12 images the surface 101 of the wafer 100 held on the holding surface 11a of the chuck table 11. The imaging unit 12 is capable of imaging the surface 101 of the wafer 100 by switching between low magnification (Lo) macro imaging and high magnification (Hi) micro imaging. The area portion (street) to be cut is detected based on the image information (key pattern) acquired by the imaging unit 12, and is used for positioning of the processing operation by the processing unit 13. The imaging unit 12 may include, for example, a light source that irradiates illumination light onto the surface of the wafer 100 held on the holding surface 11a of the chuck table 11. This light source can be composed of an epi-light source that illuminates the wafer 100 from directly above and an oblique light source that illuminates the wafer 100 from an oblique direction.

The processing unit 13 forms a cutting groove (kerf) by cutting the wafer 100 held on the holding surface 11a of the chuck table 11 along a street 103 using a rotating ring-shaped ultra-thin cutting blade 24. . The imaging unit 12 is integrated by being attached and supported by a part of the housing 25 for the processing unit 13, and can be moved in the Z-axis direction by a Z-axis moving means 15 such as a ball screw, nut, pulse motor 26, etc. It is set in. Further, the Y-axis moving means 27 for moving the imaging unit 12 and the processing unit 13 in the Y-axis direction relative to the holding surface 11a of the chuck table 11 includes a ball screw 28, a nut, a pulse motor 29, etc. Together with the shaft moving means 22, the driving means 14 is configured.

The processing unit 13 has a spindle (not shown) housed in a housing 25, and is rotatably supported by an air bearing. The spindle is rotationally driven by a motor (not shown) housed in the housing 25. The cutting blade 24 is removably attached to the tip of the spindle, and is rotated by rotation of the spindle.

Under the control of the control unit 40, the touch panel 17 displays an image of the surface 101 of the wafer 100 captured by the imaging unit 12 and various information necessary for processing, and also receives input operations necessary for processing from the operator. . The touch panel 17 is disposed at a location on the housing of the processing device 10 where it is easy to see and operate. The touch panel 17 includes a display device and an input device. The touch panel 17 may be configured with a touch screen display including a display device such as a liquid crystal display or an organic EL display, and a touch screen for specifying input positions and coordinates on the display surface of the display device.

The LED light 18 notifies the operating status of the processing device 10 using a plurality of display colors. The LED light 18 is configured to be able to turn on or blink under the control of the control unit 40.

The processing apparatus 10 configured as described above is configured to move the chuck table 11 relative to the processing unit 13 in the X-axis direction using the The machining feed is made relative to the Thereby, the streets 103 on the wafer 100 can be cut to form cutting grooves (kerfs). Subsequently, the processing apparatus 10 rotates the wafer 100 by 90 degrees by rotating the chuck table 11, and then repeats the same cutting process in the processing unit 13 for all streets 103 newly arranged in the X-axis direction. This allows the wafer 100 to be divided into individual devices 104.

The storage 30 shown in FIG. 3 stores a control program 201 that implements functions such as various processes executed by the control unit 40, data used for processing by the control program, and the like. In this embodiment, the storage 30 further stores a learned program 202 that has undergone machine learning. The storage 30 can be implemented using an HDD (Hard Disk Drive), a semiconductor memory, or the like. The storage 30 may also be used as a temporary work area when a processor included in the control unit 40 executes instructions written in the control program 201.

The storage 30 includes an image storage section 31. The image storage unit 31 stores image data 300 in which images of the front surface 101 of the wafer 100 are labeled as OK images and NG images. The image data 300 is used as teacher data for the learned program 202. The learned program 202 uses a plurality of image data 300 as training data, including an OK image in which the street 103 is located at the center of an image of the surface 101 of the wafer 100 and an NG image in which the street 103 is not located at the center of the image. This is the program I learned. The learned program 202 includes a machine-learning program for at least one of each type of wafer 100 and each type of device 104. The image data 300 includes an image taken of the surface 101 of the wafer 100 and information labeling the image into an OK image and an NG image. The image data 300 includes, for example, information indicating an image and a label. The image data 300 includes, for example, an image captured by the processing device 10 and an image acquired from outside the processing device 10.

In this embodiment, a case will be described in which the processing device 10 stores the learned program 202 in the storage 30, but the processing device 10 is not limited to this. The processing device 10 may be configured to use a learned program 202 stored in, for example, an external server, a storage device, another processing device 10, or the like that is accessible to the processing device 10. The processing device 10 may, for example, execute the learned program 202 on a server or the like.

FIG. 4 is a diagram for explaining an example of the learned program 202 used by the processing apparatus 10 according to the embodiment. In the example shown in FIG. 4, detailed description of the device 104 of the wafer 100 is omitted in the image data 300.

As shown in FIG. 4, the trained program 202 is a program (machine learning data) that has been trained using image data 300 having OK images and image data 300 having NG images as training data for each type of a plurality of wafers 100. It has become. The learned program 202 includes identification data that can identify the type of wafer 100, and a program learned corresponding to the identification data. That is, in the learned program 202, the image data 300 of OK images and NG images linked to identification data serves as teacher data on which machine learning has been performed. Machine learning may be performed using only training data of wafers 100 of the same type as the wafer 100 to be processed, and the created learned program 202 may be used during processing. The processing apparatus 10 can make the determination with high accuracy by making the determination using the learned program 202 using the same wafer 100.

Furthermore, regardless of the type of device 104, the processing apparatus 10 may perform machine learning using training data of a plurality of types of wafers 100, and may use the created learned program 202 during processing. In the latter case, linking with identification data is not essential. In this case, the processing apparatus 10 can be used for various types of wafers 100 by creating one learned program 202, so it is highly versatile and there is no need to create a learned program 202 for each type of wafer 100. Efficient.

In the example shown in FIG. 4, when the identification data is "0001" to "0004", the learned program 202 performs machine learning on a plurality of image data 300 obtained by capturing the surface 101 of the wafer 100. In detail, the trained program 202 performs machine learning on a model using a plurality of image data 300 indicating the characteristics of an OK image and a plurality of image data 300 indicating the characteristics of an NG image as a data set. This makes it possible to determine whether the image is an OK image or an NG image.

In this embodiment, the trained program 202 performs machine learning on a model using a plurality of image data 300 as a dataset, and determines whether the input image to be determined is an OK image based on the machine learned model. This is a program that can determine whether an image exists or is an NG image.

Returning to FIG. 3, the control unit 40 includes an arithmetic processing device such as a CPU (Central Processing Unit), a storage device such as a ROM (Read Only Memory) or a RAM (Random Access Memory), and an input/output interface device. The control unit 40 is a computer that can execute a control program and the like for controlling each of the above-mentioned components according to a series of processing steps performed by the processing apparatus 10 using such a device.

The control unit 40 controls the overall operation of the processing apparatus 10 according to processing conditions set by the operator via the touch panel 17. As shown in FIG. 3, the control unit 40 includes a machine learning section 41, an input section 42, a control section 43, a labeling section 44, a teaching section 45, and an alignment section 46. The control unit 40 executes the control program 201 to realize the functions, actions, etc. of each part.

The machine learning unit 41 provides a function of processing an input image using a learned program 202 and determining whether the input image is an OK image or an NG image. The machine learning unit 41 judges the input image using the learned program 202 by executing the learned program 202 that has learned teacher data including OK images and NG images, and generates information indicating the judged result. The information is provided to the control unit 43. In the present embodiment, the machine learning unit 41 executes the learned program 202 that has learned teacher data including OK images and NG images during teaching work, for example.

The machine learning unit 41 provides a function of generating and updating a learned program 202 that has undergone machine learning using the image data 300 of the image storage unit 31.

The input unit 42 inputs to the machine learning unit 41 an image captured by locating a portion of the wafer 100 to be processed placed on the chuck table 11 that is assumed to be a planned dividing line at the center of the imaging unit 12 . The locations assumed to be the planned dividing line include, for example, locations on the wafer 100 assumed (recognized) by the operator as the planned dividing lines, locations on the wafer 100 assumed to be the planned dividing lines through image recognition, and the like.

When the machine learning unit 41 determines that the image is OK, the control unit 43 determines the street value from the key pattern as the distance from the key pattern to the planned dividing line, which is used during auto-alignment to detect the scheduled dividing line of the wafer 100 to be processed. 103 distance is registered. Further, the control unit 43 provides a function of executing a warning process when the machine learning unit 41 determines that the image is an NG image. Warning processing includes, for example, displaying on the touch panel that the location of Street 103 is incorrect as the planned dividing line, notifying with audio or error sound, and lighting up the LED light installed at the top of the device in the color corresponding to the error. or notification processing performed through means such as blinking, and processing for stopping the teaching operation. If a warning process occurs, the operator performs the teaching operation again.

The labeling unit 44 labels the captured image of the front surface 101 of the wafer 100 into an OK image and an NG image, and stores them in the image storage unit 31 as image data 300. The labeling unit 44 labels an OK image in which the scheduled dividing line is positioned at the center of the imaging unit 12 and an NG image in which the scheduled dividing line is not positioned at the center of the imaging unit 12, and stores the labels as image data 300 in the image storage unit 31. to be memorized. The labeling unit 44 acquires, for example, an image captured by the processing device 10, an image captured by another processing device 10, etc., and associates a label indicating an OK image or a NG image with the image. For example, the labeling unit 44 may be configured to display a screen on the touch panel 17 that allows the user to select whether the image is an OK image or an NG image, and label the image based on the operation result on the screen. . The labeling unit 44 determines the planned dividing line in the image, for example, if there is a straight line in the central area of the screen divided into three, the planned dividing line is positioned at the center and the image is determined to be OK. A configuration may also be adopted in which labels are automatically linked based on conditions or the like.

The teaching unit 45 provides a function of registering the key pattern formed on the surface 101 of the wafer 100 and the distance from the key pattern to the street 103 as the distance between the key pattern and the planned dividing line. The teaching unit 45 causes the touch panel 17 to display a screen for registering the distance from the key pattern that is a part of the device 104 formed on the surface 101 of the wafer 100 to the street 103, and displays the key pattern registered via the touch panel 17. and the distance to the street 103 are made to correspond to the type of wafer 100, etc.

In this embodiment, when the machine learning section 41 determines that the image is OK, the teaching section 45 registers the key pattern and the distance of the street 103 from the key pattern, and when the machine learning section 41 determines that the image is NG, Provides the ability to issue a warning. That is, the teaching unit 45 executes processing according to the determination result of the machine learning unit 41 on the image registered by the operator in the teaching operation. The teaching unit 45 may be configured to register information regarding the street 103 in addition to the key pattern and the distance of the street 103 from the key pattern. The teaching unit 45 may register information regarding the position of the street 103, for example. The information regarding the position of the street 103 includes, for example, information that allows identification of the street 103 in the image, such as the coordinate position and width of the street 103.

The alignment unit 46 detects the streets 103 of the wafer 100 by capturing an image of the wafer 100 to be processed placed on the chuck table 11 and detecting the key pattern by pattern matching.

The configuration example of the processing apparatus 10 according to the present embodiment has been described above. Note that the above configuration described using FIGS. 1 to 4 is just an example, and the configuration of the processing apparatus 10 according to the present embodiment is not limited to the example. The functional configuration of the processing device 10 according to this embodiment can be flexibly modified according to specifications and operation.

(Example of image data registration)
Next, an example of registration of the image data 300 executed by the processing device 10 according to the embodiment will be described. FIG. 5 is a flowchart illustrating an example of a processing procedure regarding registration of image data 300 according to the embodiment. The processing procedure shown in FIG. 5 is realized by the control unit 40 executing the control program 201. The processing procedure shown in FIG. 5 is executed, for example, at the timing when an image is registered as teacher data, at the timing when an image to be registered in the learned program 202 is generated, and the like.

As shown in FIG. 5, the control unit 40 acquires an image of the surface 101 of the wafer 100 (step 1001). The control unit 40 acquires, for example, information on an image of the device 104 on the front surface 101 of the wafer 100. For example, when determining accuracy, the control unit 40 acquires images for each type of device 104. For example, when seeking versatility, the control unit 40 acquires images of not only one type of device 104 but also an unspecified number of devices 104. For example, the control unit 40 may be configured to acquire images not only from the imaging unit 12 but also from a server, a storage device, etc. external to the processing device 10. When the process of step 1001 is completed, the control unit 40 advances the process to step 1002.

The control unit 40 stores image data 300 obtained by labeling the acquired image into an OK image and an NG image in the image storage unit 31 (step 1002). For example, the control unit 40 can be configured to label OK images and NG images based on the arrangement and arrangement conditions of the streets 103 in the image. For example, the control unit 40 can be configured to display on the touch panel 17 a screen that allows the user to determine whether the image is OK or not, and to perform labeling based on the determination result. When the process of step 1002 is completed, the control unit 40 advances the process to step 1003.

The control unit 40 performs machine learning using the labeled image data 300 and generates a trained program 202 that can discriminate between OK images and NG images (step 1003). For example, the control unit 40 has learned a rule (model) for determining unknown images for each wafer 100 and device 104 based on the characteristics, patterns, etc. of OK images and NG images of the image data 300 that is teacher data. A learned program 202 is generated. That is, generating the learned program 202 means, for example, that a rule for determining an unknown image has been learned. When the process of step 1003 is completed, the control unit 40 advances the process to step 1004.

The control unit 40 stores the generated learned program 202 in the storage 30 (step 1004). For example, when the learned program 202 is already stored in the storage 30, the control unit 40 overwrites the generated learned program 202 or stores it so as to add a new one. When the control unit 40 finishes the process of step 1004, it ends the process procedure shown in FIG.

By executing the processing procedure shown in FIG. 5, the control unit 40 functions as a machine learning section 41 and a labeling section 44.

(Judgment by machine learning department)
Next, an example of the determination by the machine learning unit 41 executed by the processing apparatus 10 according to the embodiment will be described. FIG. 6 is a flowchart illustrating an example of a processing procedure regarding determination by the machine learning unit 41 according to the embodiment. The processing procedure shown in FIG. 6 is realized by the control unit 40 executing the control program 201. The processing procedure shown in FIG. 6 is executed, for example, when an image is input from the input unit 42.

As shown in FIG. 6, the control unit 40 processes the input image using the trained program 202, and determines whether the input image is an OK image or an NG image (step 1101). For example, by executing the learned program 202, the control unit 40 causes the learned program 202 to determine whether the input image is an OK image or an NG image. Upon acquiring the determination result from the learned program 202, the control unit 40 advances the process to step 1102.

The control unit 40 outputs the determination result of the input image to the control section 43 (step 1102). When the control unit 40 outputs the determination result of the input image obtained by executing the learned program 202 to the control unit 43, it ends the processing procedure shown in FIG. 6.

In this embodiment, the control unit 40 functions as the machine learning section 41 by executing the processing procedure shown in FIG.

(Teach section processing)
Next, an example of the process of the teaching unit 45 executed by the processing apparatus 10 according to the embodiment will be described. FIG. 7 is a flowchart illustrating an example of a processing procedure regarding the teaching unit 45 according to the embodiment. The processing procedure shown in FIG. 7 is realized by the control unit 40 executing the control program 201. The processing procedure shown in FIG. 7 is executed, for example, when the processing device 10 performs a teaching operation.

As shown in FIG. 7, the control unit 40 acquires an image of the surface 101 of the wafer 100 (step 2001). The control unit 40 selects a characteristic pattern as a key pattern in the device region formed on the surface 101 of the wafer 100 (step 2002). In auto alignment, which images the wafer 100 before processing and detects the position of the planned dividing line, the position of the street 103 is automatically determined using the key pattern registered in the teaching section 45 as a clue.

FIG. 8 is a diagram for explaining an example of registration of key patterns and streets according to the embodiment. As shown in FIG. 8, the control unit 40 allows the operator to select a target key pattern 120 from the image. Next, the control unit 40 controls the driving means to position the street 103 at the center of the imaging unit in the Y direction. After that, when the registration button for registering the street 103 as the position of the planned dividing line is operated, the control unit 40 recognizes the distance 130 from the selected key pattern 120 to the street 103, and the key pattern 120 and the distance 130 are Link and register. For example, the control unit 40 detects the distance by aligning the key pattern 120 with the center of the image of the imaging unit 12 in the Y-axis direction and then moving the street 103 to the center of the image in the Y-axis direction.

In the example shown in FIG. 8, images 301 and 302 are images obtained by enlarging a part of the surface 101 of the wafer 100 at high magnification. A plurality of patterns 121 are located along the line 120. The operator selects the characteristic key pattern 120 as a target in the image 301, and performs a teaching operation to select and register the street 103. In this case, the control unit 40 detects the distance 130 from the selected key pattern 120 to the street 103, and registers the key pattern 120 and the distance 130 in association with each other.

However, as shown in the image 302, if the street 103 and the device pattern 121 are similar, the operator may mistakenly recognize the pattern 121 similar to the street 103 as the planned dividing line during the teaching operation, and It may be set as . Furthermore, due to an operational error, there is also a possibility that a device area other than the street 103 may be positioned at the center of the imaging unit and the registration button may be pressed. For example, the operator selects the characteristic key pattern 120 as a target in the image 302, and mistakenly selects and registers the pattern 121 as the street 103. In this case, the control unit 40 detects the distance 131 from the selected key pattern 120 to the pattern 121, and registers the key pattern 120 and the incorrect distance 131 in association with each other. As a result, the conventional processing apparatus may continue to process the position of the wafer 100 that has been incorrectly registered by the operator.

In the present invention, a learned program 202 created based on images of the type of wafer 100 to be processed or a learned program 202 created based on images of various types of wafers 100 is stored. Therefore, when an operator recognizes a device area similar to Street 103 as a scheduled dividing line and inputs an image captured by positioning it at the center to the machine learning unit, the color and pattern of the pattern will be different from the normal scheduled dividing line. Since the scheduled dividing lines stored in the learned program 202 are not similar to the OK image at the center, there is a high possibility that the image will be determined as an NG image. Therefore, the processing apparatus 10 prevents the operator from registering the wrong position of the wafer 100 as the street 103 during the teaching operation.

Returning to FIG. 7, the control unit 40 controls the driving means 14 so that the planned dividing line is located at the center of the imaging unit 12 (step 2003). When the planned dividing line of the wafer 100 is located at the center of the imaging unit 12, the control unit 40 advances the process to step 2004. Note that the center of the imaging unit 12 when locating the location assumed to be the planned dividing line is the center in the Y direction, which is the indexing feed direction that intersects the X direction, which is the processing feed direction in which the planned dividing line extends. be.

The control unit 40 determines whether a registration operation is detected (step 2004). For example, when the control unit 40 displays a screen for registering the street 103 on the touch panel 17 and detects a registration operation on the screen via the touch panel 17, it determines that a registration operation has been detected. If the control unit 40 determines that no registration operation has been detected (No in step 2004), the control unit 40 determines not to register, and ends the processing procedure shown in FIG. 7. Further, if the control unit 40 determines that a registration operation has been detected (Yes in step 2004), the control unit 40 advances the process to step 2005.

The control unit 40 inputs the image being captured by the imaging unit 12 to the machine learning section 41 (step 2005). For example, the control unit 40 inputs the captured image to the machine learning unit 41, so that the machine learning unit 41 executes the learned program 202 and determines whether the image is an OK image or an NG image. be done. Upon acquiring the determination result output by the machine learning section 41, the control unit 40 advances the process to step 2006.

The control unit 40 determines whether the determination result is an OK image (step 2006). For example, when the determination result output by the machine learning section 41 indicates an OK image, the control unit 40 determines that the determination result is OK. When the control unit 40 determines that the image is an OK image (Yes in step 2006), the control unit 40 advances the process to step 2007.

The control unit 40 registers the key pattern 120 of the wafer 100 and the distance 130 from the key pattern 120 to the street 103 (step 2007). For example, the control unit 40 registers the center position of the imaging unit 412 as the street 103, and registers the street 103 and the distance 130 in association with each other. That is, if the control unit 40 attempts to register a location that is not the street 103 as the street 103, the learned program 202 determines that it is an NG image, so the control unit 40 registers the street 103 without registering the incorrect distance 131 described above. It can be registered in association with the distance 130. When step 2007 is completed, the control unit 40 ends the processing procedure shown in FIG. 7 .

Further, when the control unit 40 determines that the image is not an OK image (No in step 2006), the control unit 40 advances the process to step 2008. The control unit 40 executes the above warning process (step 2008). For example, by executing a warning process, the control unit 40 warns that the position of the street 103 is incorrect, stops registration in the teaching operation, or causes the street 103 to be re-registered. Examples of means for giving a warning include displaying a message indicating a warning on the touch panel 17, lighting or blinking an LED light 18 installed at the top of the device in a color corresponding to the error, and emitting a voice or error sound indicating a warning. can give. When the process ends in step 2008, the control unit 40 ends the process procedure shown in FIG.

In this embodiment, the control unit 40 functions as the teaching section 45 by executing the processing procedure shown in FIG.

As explained above, in the teaching operation, the processing device 10 uses the learned program 202 to determine whether the captured image of the street 103 of the wafer 100 is an OK image or an NG image. If it is determined that there is, a warning process is executed. When the learned program 202 determines that the image is an OK image, the processing apparatus 10 accepts registered contents such as the street 103 of the wafer 100 registered by the operator. As a result, the processing device 10 uses the learned program 202 that has learned the teacher data including OK images and NG images actually captured to check the validity of the street 103, distance 130, etc. registered by the operator. It can be confirmed based on the image. As a result, the processing apparatus 10 can improve the accuracy of registering the planned dividing line and prevent continued processing of incorrect positions on the wafer 100.

Furthermore, the processing apparatus 10 can cause the learned program 202 to learn image data 300 obtained by labeling an image taken during a teaching operation of the front surface 101 of the wafer 100 into an OK image and an NG image as new teacher data. Thereby, the processing device 10 can improve the accuracy of determining the learned program 202 as the learned program 202 is used. As a result, the processing apparatus 10 can prevent processing the wafer 100 at an incorrect position that is not the planned dividing line with even higher accuracy.

(Outline of processing during auto alignment)
Next, an example of a process for preventing erroneous machining during auto-alignment performed by the machining apparatus 10 according to the embodiment will be described. FIG. 9 is a flowchart illustrating an example of a processing procedure during auto-alignment of the processing apparatus 10 according to the embodiment. The processing procedure shown in FIG. 9 is realized by the control unit 40 executing the control program 201. The processing procedure shown in FIG. 9 is executed while the chuck table 11 on which the wafer 100 is placed is placed at the imaging position of the imaging unit 12.

As shown in FIG. 9, the control unit 40 images the wafer 100 to be processed using the imaging unit 12 (step 3001). For example, the control unit 40 causes the imaging unit 12 to image the surface 101 of the wafer 100, and acquires the captured image. When the process of step 3001 is completed, the control unit 40 advances the process to step 3002.

Control unit 40 detects key pattern 120 from the image by pattern matching (step 3002). The control unit 40 detects a key pattern 120 that matches or is similar to a previously registered key pattern, for example, by analyzing the acquired image. When the process of step 3002 is completed, the control unit 40 advances the process to step 3003.

Control unit 40 detects street 103 of wafer 100 based on key pattern 120 (step 3003). For example, the control unit 40 detects a position separated by a distance 130 from the key pattern 120 as the street 103 based on the key pattern 120 and the distance 130 associated with the key pattern 120. Upon completion of the process in step 3003, the control unit 40 advances the process to step 3004.

The control unit 40 positions the detected street 103 at the center of the imaging unit 12 and images it (step 3004). The control unit 40 controls, for example, to drive the driving means 14 so that the detected street 103 is located at the center of the imaging unit 12, and causes the imaging unit 12 to image the street 103 of the wafer 100. When the process of step 3004 is completed, the control unit 40 advances the process to step 3005.

The control unit 40 inputs the captured image to the machine learning section 41 (step 3005). For example, the control unit 40 inputs the captured image to the machine learning unit 41, so that the machine learning unit 41 executes the learned program 202 and determines whether the image is an OK image or an NG image. Ru. Upon acquiring the determination result output by the machine learning section 41, the control unit 40 advances the process to step 3006.

The control unit 40 determines whether the determination result is an OK image (step 3006). For example, when the determination result output by the machine learning section 41 indicates an OK image, the control unit 40 determines that the determination result is OK. When the control unit 40 determines that the image is an OK image (Yes in step 3006), the control unit 40 advances the process to step 3007.

The control unit 40 controls processing of the wafer 100 (step 3007). The control unit 40 forms cutting grooves on the wafer 100 while controlling the cutting depth by controlling the cutting blade 24 to move up and down depending on the height of the upper surface of the street 103, for example. More specifically, the control unit 40 causes the driving means 14 to move the cutting blade 24 along the Y-axis direction, and the Z-axis moving means 15 to move the cutting blade 24 along the Z-axis direction to a position corresponding to the top surface height. A cut groove is formed in the street 103 by positioning the groove. When the process of step 3007 is completed, the control unit 40 ends the processing procedure shown in FIG. 9 .

Further, when the control unit 40 determines that the image is not an OK image (No in step 3006), the control unit 40 advances the process to step 3008. The control unit 40 executes the above warning process (step 3008). For example, the control unit 40 executes a warning process to warn that the position of the street 103 is incorrect, and causes the processing unit 13 to stop or suspend processing. When the process of step 3008 is completed, the control unit 40 ends the processing procedure shown in FIG. 9 .

In the processing procedure shown in FIG. 9, the control unit 40 functions as the alignment section 46 by executing a series of processing from step 3001 to step 3004. The control unit 40 functions as the input unit 42 by executing the process of step 3005. The control unit 40 functions as the control section 43 by executing the processes from step 3006 to step 3008.

In addition, since the learned program is used to determine the street when detecting streets in the teaching operation, the processes using the learned program from steps 3004 to 3008 in FIG. 9 in FIG. 9 do not need to be performed during auto-alignment. good.

As explained above, the processing apparatus 10 uses the learned program 202 during processing to determine whether the captured image of the street 103 of the wafer 100 is an OK image or an NG image, and determines whether the image is an NG image. If it is determined, a warning process is executed. The processing apparatus 10 processes the wafer 100 when the learned program 202 determines that the image is OK. Thereby, the processing device 10 can improve the image determination accuracy by using the determination result of the learned program 202 that has learned the teacher data including the actually captured OK image and NG image. As a result, the processing apparatus 10 can prevent processing the wrong position of the wafer 100 that is not the street (planned dividing line) 103 when the planned dividing line is not captured at the center of the image. Furthermore, by using the learned program 202 that has learned actual images, the processing apparatus 10 can make appropriate determinations for each wafer 100 and device 104, thereby improving convenience.

Further, the processing apparatus 10 can cause the learned program 202 to learn image data 300 obtained by labeling an image of the surface 101 of the wafer 100 into an OK image and an NG image as teacher data. Thereby, the processing device 10 can improve the determination accuracy of the learned program 202 with respect to the image input to the learned program 202. As a result, the processing apparatus 10 can more reliably prevent processing at an incorrect position on the wafer 100 that is not on the planned dividing line.

Further, the processing device 10 images the wafer 100 to be processed placed on the chuck table 11, detects the key pattern 120 by pattern matching, thereby detecting the street 103 of the wafer 100, and detects the street 103 of the wafer 100. can be positioned at the center of the imaging unit 12 and imaged. Thereby, the processing device 10 can improve the accuracy of the image itself input to the learned program 202. As a result, the processing apparatus 10 can more reliably prevent processing at an incorrect position on the wafer 100 that is not on the planned dividing line. For example, by applying the processing apparatus 10 to auto-alignment, even if the position of the street 103 is automatically determined using the key pattern as a clue, processing the wafer 100 at the wrong position can be prevented.

Although the processing apparatus 10 according to the above embodiment executes the learned program 202 during teaching work and processing to determine whether an image is an OK image or an NG image, the present invention is not limited to this. Not done. For example, the processing device 10 may be configured to execute the learned program 202 only during a teaching operation. Thereby, the machining device 10 may improve the possibility of preventing machining at an incorrect position during machining by improving the accuracy of registering the dividing line in the teaching operation.

[Modified example]
A processing apparatus 10 according to a modification of the above embodiment will be described below. FIG. 10 is a perspective view schematically showing a configuration example of a processing device 10 according to a modification of the embodiment. FIG. 11 is a block diagram showing a configuration example of the laser irradiation unit shown in FIG. 10. In addition, in each of the following embodiments, the same portions are given the same reference numerals and redundant explanations will be omitted.

A processing apparatus 10 shown in FIG. 10 is a laser processing apparatus that irradiates a wafer 100 with a laser beam. The processing apparatus 10 basically includes a chuck table 11 that holds the wafer 100, an imaging unit 12, a processing unit 13, a driving means 14, a Z-axis moving means 15, a touch panel 17, and a control unit 40. The processing unit 13 includes a laser irradiation unit 5 instead of the cutting blade 24 described above.

As shown in FIG. 11, the laser irradiation unit 5 includes an oscillator 51, an intensity adjustment section (attenuator) 52, a polarization direction setting section 53, a mirror element 54, and a condenser lens 55. The laser irradiation unit 5 is controlled by a control unit 40.

The oscillator 51 is for oscillating a laser beam of a predetermined wavelength, and is configured by, for example, a laser beam oscillator such as a YAG laser oscillator or a YVO4 laser oscillator, an oscillator that oscillates a laser beam in an infrared wavelength range such as 1064 nm, or the like. The intensity adjustment section 52 includes a 1/2 wavelength plate 521, a motor 522, a polarizing beam splitter 523, and an absorber 524.

The half-wave plate 521 is arranged downstream of the oscillator 51 so as to be rotatable by a motor 522 . The half-wave plate 521 changes the linear polarization direction of the laser beam LB emitted by the oscillator 51 according to its rotation angle. The polarizing beam splitter 523 transmits a laser beam having a predetermined linear polarization direction among the laser beams that have passed through the half-wave plate 521 toward the polarization direction setting unit 53, and transmits a laser beam having a predetermined linear polarization direction other than the predetermined linear polarization direction. The laser beam having the above is branched to the absorber 524 side. The absorber 524 is for absorbing the laser beam split by the polarizing beam splitter 523, and is preferably made of, for example, a matte black metal to absorb the laser beam well.

The polarization direction setting section 53 includes a 1/2 wavelength plate 531 and a motor 532. The 1/2 wavelength plate 531 is rotatably arranged by a motor 532 after the polarizing beam splitter 523 . The half-wave plate 531 changes the linear polarization direction of the laser beam transmitted through the polarizing beam splitter 523 according to its rotation angle. The polarization direction setting section 53 functions as a polarization direction setting means according to the present invention. The mirror element 54 is for reflecting the laser beam that has passed through the 1/2 wavelength plate 531 toward the condenser lens 55 side. The condensing lens 55 is disposed to face the wafer 100, locates a condensing point inside the wafer 100, and condenses the laser beam reflected by the mirror element 54.

In the example shown in FIG. 11, the condenser lens 55 functions as a condenser, for example. The intensity adjustment section 52 and the polarization direction setting section 53, which are arranged between the oscillator 51 and the condensing lens 55, function as, for example, an optical system.

The configuration example of the processing device 10 according to the modified example of this embodiment has been described above. Note that the above configuration described using FIG. 8 is just an example, and the configuration of the processing apparatus 10 according to the present embodiment is not limited to this example. The functional configuration of the processing device 10 according to this embodiment can be flexibly modified according to specifications and operation.

A processing apparatus 10 according to a modification of the embodiment processes a wafer 100 partitioned by streets 103, which are planned dividing lines, along streets 103. For example, the processing apparatus 10 irradiates the wafer 100 with a laser beam of a transparent wavelength to continuously form a modified layer (altered region) along the streets 103 inside the wafer 100 . Furthermore, the processing device 10 may have a function of dividing the wafer 100 by applying an external force along the streets 103 whose strength has been reduced due to the formation of the modified layer. Further, the processing apparatus 10 irradiates the wafer 100 with a laser beam having an absorbing wavelength, for example, to form cutting grooves along the streets 103 on the surface of the wafer 100. The cutting groove may have a depth to half-cut the wafer 100, or may have a depth to fully cut the wafer 100.

The processing apparatus 10 can use the processing procedure shown in FIGS. 7 and 9 described above. For example, the processing apparatus 10 may replace the processing in step 3007 shown in FIG. 9 with processing performed by the laser irradiation unit 5.

For example, if the control unit 40 of the processing apparatus 10 determines in step 3005 that the image is an OK image (Yes in step 3005), it controls the processing of the wafer 100 (step 3007). For example, the control unit 40 positions one end of the street 103 vertically below the condenser lens 55 and moves the chuck table 11 relative to the processing unit 13 to the X-axis moving means 22 while the laser beam LB is emitted from the oscillator 51. to relatively feed processing in the X-axis direction. When the process of step 3007 is completed, the control unit 40 ends the processing procedure shown in FIG. 9 .

Further, when the control unit 40 determines that the image is not an OK image (No in step 3006), the control unit 40 advances the process to step 3008. The control unit 40 executes the above warning process (step 3008). For example, the control unit 40 executes a warning process to warn that the position of the street 103 is incorrect, and causes the processing unit 13 to stop or suspend processing. When the process of step 3008 is completed, the control unit 40 ends the processing procedure shown in FIG. 9 .

In the teaching operation, the processing apparatus 10 according to the modified example of the embodiment uses the learned program 202 to determine whether the captured image of the street 103 of the wafer 100 is an OK image or an NG image, and determines whether the image taken of the street 103 of the wafer 100 is an OK image or an NG image. If it is determined that this is the case, a warning process is executed. When the learned program 202 determines that the image is an OK image, the processing apparatus 10 accepts registered contents such as the street 103 of the wafer 100 registered by the operator. As a result, the processing device 10 uses the trained program 202 that has learned training data including OK images and NG images that were actually captured to check the validity of the street 103, distance 130, etc. registered by the operator. It can be confirmed based on the image. As a result, the processing apparatus 10 can improve the accuracy of registering the planned dividing line and prevent continued processing of incorrect positions on the wafer 100.

Further, the processing apparatus 10 can cause the trained program 202 to learn image data 300 obtained by labeling an image of the front surface 101 of the wafer 100 during a teaching operation into an OK image and an NG image as teacher data. Thereby, the processing device 10 can improve the determination accuracy of the learned program 202 with respect to the image input to the learned program 202. As a result, the processing apparatus 10 can more reliably prevent processing at an incorrect position on the wafer 100 that is not on the planned dividing line.

Furthermore, similarly to the embodiment, the processing apparatus 10 according to the modification of the embodiment uses the learned program 202 to determine whether the image taken of the street 103 of the wafer 100 is an OK image or an NG image. However, if it is determined that the image is an NG image, a warning process is executed. The processing apparatus 10 processes the wafer 100 when the learned program 202 determines that the image is OK. Thereby, the processing device 10 can improve the image determination accuracy by using the determination result of the learned program 202 that has learned the teacher data including the actually captured OK image and NG image. As a result, when the planned dividing line is not captured at the center of the image, the processing apparatus 10 can prevent processing at an incorrect position of the wafer 100 that is not at the planned dividing line. Furthermore, by using the learned program 202 that has learned actual images, the processing apparatus 10 can make appropriate determinations for each wafer 100 and device 104, thereby improving convenience.

Note that the present invention is not limited to the above embodiments and modifications. That is, various modifications can be made without departing from the gist of the invention.

For example, in the above embodiments and modifications, the processing apparatus 10 has been described with reference to the case where the control unit 40 includes the machine learning section 41, the input section 42, the control section 43, the labeling section 44, the teaching section 45, and the alignment section 46. , but not limited to. For example, the processing device 10 may not include the labeling section 44 and the alignment section 46, or may include either the labeling section 44 and the alignment section 46.

In the above embodiments and modified examples, the learned program 202 is created by performing machine learning for each type of wafer 100, but the learned program 202 is not limited to this. The learned program 202 may be created by performing machine learning on a plurality of types of wafers 100. As a result, the learned program 202 is a general-purpose program obtained by machine learning various types of wafers 100, so that it is not necessary to create it for each type of wafer 100, and the versatility can be increased.

10 processing device 11 chuck table 12 imaging unit 13 processing unit 17 touch panel 18 LED light 30 storage 31 image storage section 40 control unit 41 machine learning section 42 input section 43 control section 44 labeling section 45 teaching section 46 alignment section 100 wafer (workpiece) thing)
103 Street (planned dividing line)
104 Device 201 Control program 202 Learned program 300 Image data (teacher data)
301, 302 images

Claims (4)

A processing device for processing a workpiece formed on a surface in which a device region is divided by a plurality of planned dividing lines,
The processing equipment is
a holding table that holds the workpiece;
a processing feed unit that processes and feeds the holding table in the X direction;
a processing unit that forms a processing groove in a workpiece held on the holding table;
an indexing feed unit that indexes and feeds the processing unit in a Y direction intersecting the X direction;
an imaging unit that images the workpiece held on the holding table;
a control unit;
Equipped with
The control unit includes:
A trained program that has learned teacher data having an OK image in which the planned dividing line is located at the center of the image of the surface of the workpiece, and an NG image in which the scheduled dividing line is not located in the center of the image, a machine learning unit that processes the input image and determines whether the input image is an OK image or an NG image;
an input unit that inputs an image captured by positioning a location of the workpiece assumed to be the planned dividing line at the center of the imaging unit to the machine learning unit;
a teach unit that registers a key pattern formed on the surface of the workpiece;
Equipped with
The teaching section is
If the machine learning unit determines that the image input to the machine learning unit by the input unit is an OK image, registering the key pattern and the distance of the planned dividing line from the key pattern;
A processing device characterized by issuing a warning when the machine learning unit determines that the image is an NG image. The control unit includes:
An image in which the planned division line is positioned at the center of the imaging unit is labeled as an OK image and an NG image in which the planned division line is positioned off the center of the imaging unit, and the labels are stored in the image recording unit as image data. Further equipped with a labeling section,
The machine learning unit uses a trained program that has learned the image data labeled by the labeling unit of the image recording unit as part of the teacher data, and determines whether the input image is an OK image or an NG image. The processing device according to claim 1, wherein the processing device determines whether there is any. The processing unit is
The processing device according to claim 1 or 2, further comprising a cutting blade held at the tip of a rotatable spindle. The processing unit is
an oscillator that emits a laser beam;
a condenser that condenses the laser beam oscillated by the oscillator and irradiates the workpiece held on the holding table;
an optical system disposed between the oscillator and the condenser and guiding the laser beam to the oscillator;
The processing device according to claim 1 or 2, characterized by comprising:

Images