TWI797617B - Mesh mask inspection device - Google Patents

Abstract

[Problem] Provide a screen mask inspection device that can suppress the occurrence of printing defects, etc. [Solution] The metal mask inspection device 30 is provided with: illuminating devices 32A, 32B, which can irradiate predetermined light on a predetermined inspected portion which is between the front and back surfaces of the metal mask 20 during solder printing. The peripheral part of the opening 21 on the back side 201 side which will be the substrate contacting side which will be in contact with the printed circuit board; the camera 32C photographs the inspected part on the back side 201 of the metal mask 20; and the control device controls the above-mentioned configuration. The control device is configured to obtain shape data related to the inspected part based on image data related to the inspected part captured by the camera 32C, and determine the shape data related to the inspected part based on the shape data related to the inspected part. Whether the inspected department is good or not.

Description

Mesh mask inspection device

The present invention relates to a mesh mask inspection device for inspecting a mesh mask used for solder printing on a substrate.

Generally, in a production line for mounting electronic components on a substrate, firstly, paste solder is printed on lands of the substrate by screen printing on a solder printing machine. Next, the electronic components are temporarily fixed on the substrate by utilizing the viscosity of the cream solder. Then, soldering is performed by introducing the aforementioned substrate into a reflow furnace.

A solder printer is provided with a mesh mask having a plurality of openings formed corresponding to the arrangement of islands of a substrate to be printed in advance. When performing solder printing, with the mesh mask in contact with the surface of the substrate, cream flux is supplied to the upper surface, and the cream flux is pushed into the openings by the sliding squeegee. Then, by separating the mesh mask from the substrate, the cream solder filled in the openings is released and printed (transferred) on the islands.

In such a solder printing machine, the screen mask is replaced when, for example, defective solder printing occurs or a specification of a substrate to be printed is changed.

Next, when solder printing starts again using the replaced mesh mask, the mesh mask may be inspected in advance in order to prevent printing defects from occurring in the first place.

The technology of implementing this kind of inspection, for example, in the state where the screen mask is installed on the solder printing machine, the technology of using the laser measuring device to measure the top of the screen mask three-dimensionally to determine whether the screen mask is good or not has been established. It is well known (see, for example, Patent Document 1). [Prior Art Literature] [Patent Document]

[Patent Document 1] Japanese Patent Laid-Open No. 2001-315310

[Problem to be solved by the invention]

However, in the structure of the above-mentioned Patent Document 1, when solder printing is performed, three-dimensional soldering is performed on the front and back sides of the mesh mask from the surface (upper) side opposite to the back (lower) side that contacts the printed circuit board. In the measurement, the back (lower surface) is the substrate abutting surface side, and the surface (upper surface) is the substrate non-contacting surface side.

Therefore, there is a possibility that various abnormalities (defective parts) on the back side of the mesh mask, which cause penetration or flow of the cream solder transferred to the island portion during solder printing, may not be detected.

For example, the corners of the openings on the back side of the net mask are excessively rounded due to excessive electrolytic polishing [see Fig. 10(b)], or there are openings on the back side of the net mask. In the case of flaws such as recesses or cuts that are connected to each other [see Fig. 10(c)], or when deformation occurs on the periphery of the opening on the back side of the mesh mask [see Fig. 10(d)], solder printing may occur. When the paste flux flows, permeates or flows from here, it may even cause poor printing such as bridging (bridge), and the soldering operation of the mounting parts cannot be properly performed, and the occurrence rate of defective products may increase.

The present invention was developed in view of the above circumstances, and an object of the present invention is to provide a screen mask inspection device that can suppress occurrence of printing defects and the like. [Means to solve the problem]

Hereinafter, each means suitable for solving the above-mentioned problems will be described item by item. In addition, if necessary, the specific function and effect of the corresponding means are also stated.

Means 1. A mesh mask inspection device having a plurality of openings for inspecting a mesh mask used for solder printing on a substrate, characterized by having: At least one irradiating means capable of irradiating a predetermined light onto a portion to be inspected that is in contact with the substrate that will be in contact with the substrate during solder printing among the front and back sides of the mesh mask The peripheral portion of the aforementioned opening on the back side of the front side; photographing means capable of photographing the portion to be inspected irradiated by the predetermined light; The shape data acquisition means can obtain the shape data related to the inspected part based on one or more image data related to the inspected part captured by the aforementioned shooting means; and The judging means judges the quality of the inspected part based on the shape data related to the inspected part acquired by the shape data acquiring means.

According to the above method 1, of the front and back surfaces of the mesh mask, inspection is performed on the portion to be inspected which is the peripheral portion of the opening from the back side of the substrate abutting surface side which is in contact with the substrate during solder printing.

With this configuration, it is possible to detect various abnormalities (defective parts) related to the periphery of the opening on the back side of the mesh mask (inspected part) which may cause penetration or flow of flux during printing. As a result, occurrence of printing defects and the like can be suppressed.

Means 2. The net mask inspection device as described in means 1, wherein the aforementioned determination means is configured to determine whether the size of the chamfered portion on the periphery of the aforementioned opening exceeds a predetermined amount (for example, the "" of the chamfered portion Whether at least one of the judging items among “depth” and “width” exceeds a predetermined critical value) to judge whether the aforementioned inspected part is good or not.

As described in the above "Problems to be Solved by the Invention", the peripheral corners of the openings on the back side of the mesh cover may become excessively rounded due to excessive electrolytic polishing [see Fig. 10(b) )], so during flux printing, there is a risk of causing flux to flow, permeate or flow from here. For this point, if according to the above-mentioned means 2, this shortcoming can be prevented from happening before it happens.

Means 3. The net mask inspection device as in means 1 or 2, wherein the aforementioned judging means is configured to transparently judge whether there is an abnormal part exceeding a predetermined size (for example, "depth", "width") in the aforementioned inspected part , "Length" and "Volume" in at least one of the judgment items whether there is an abnormal part exceeding a predetermined critical value), and judge whether the aforementioned inspected part is good or not.

As mentioned above in "Problems to be Solved by the Invention", when there is a wound such as a depression or a cut on the back side of the mesh cover that communicates with the opening [see Fig. 10(c)], the opening on the back side of the mesh cover When the peripheral edge is deformed [see FIG. 10(d)], there is a possibility that the solder paste flows from this area during flux printing, causing penetration or bleeding. For this point, if according to the above-mentioned means 3, it is possible to prevent this defect from happening before it happens. That is, the above-mentioned "abnormal part" includes wounds such as depressions or cuts communicating with the opening, and deformation of the periphery of the opening. In addition, the above-mentioned "abnormal part" may also include a part where foreign matter adheres (foreign matter adhering to the part to be inspected), and the like.

Means 4. The mesh mask inspection device according to any one of means 1 to 3, wherein a discrimination means (generated model) is provided for an encoding unit (Encoder) which extracts feature quantities from input shape data, and The neural network of the decoding unit (decoder) that reconstructs the shape data from this feature value uses only the shape data (shape data belonging to only good products) related to the above-mentioned inspected part without abnormality as learning data to generate, The aforesaid judgment means have: The reconstructed shape data obtaining means can obtain the reconstructed shape data related to the inspected part, and the reconstructed shape data related to the inspected part is the shape data related to the inspected part obtained by the aforementioned shape data obtaining means As the original shape data, it is input and reconstructed by inputting the aforementioned identification means; and comparison means, comparing the aforementioned original shape data with the aforementioned reconstructed shape data; In addition, it is configured such that the quality of the inspected part can be judged based on the comparison result obtained by the comparison means.

In addition, the above-mentioned "neural network" includes, for example, a convolutional neural network having a plurality of convolutional layers. In addition, the above-mentioned "learning" includes, for example, deep learning and the like. The aforementioned "discrimination means (generated model)" includes, for example, an autoencoder (autoencoder), a convolutional autoencoder (convolutional autoencoder), and the like.

According to the above-mentioned method 4, it is possible to use a discrimination method (generated model) such as an autoencoder constructed by a learning neural network to determine whether there is an abnormality (defective part) in the inspected part. With this configuration, minute abnormalities that were difficult to detect in the past can be detected.

Furthermore, in this method, since the original shape data captured by the inspected part is compared with the reconstructed shape data reconstructed based on the original shape data, there is no object belonging to the inspection object among the compared two shape data. Influences caused by differences in imaging conditions on the side of the net mask (for example, arrangement position or arrangement angle of the net mask, bending, etc.), or differences in imaging conditions on the side of the inspection device (for example, lighting conditions or camera angles, etc.), Finer abnormalities can be detected more accurately.

In addition, if the inspection is carried out on the part to be inspected (periphery of the opening) at the predetermined position of the net cover, the composition information (position data or size data, etc.) of the part to be inspected must be used as a criterion for judging whether it is good or not In the composition of the standard, for example, the substrate design information such as Jaber data is stored in advance, and the composition information of the inspected part to be inspected is obtained appropriately, and compared with the composition information, the information of the inspected part to be inspected is Good or bad is judged, so there is a risk of lower inspection efficiency. Furthermore, the positioning of the mesh mask to the inspection position must also be done correctly.

Regarding this point, if according to this method, it is the structure of inspecting each inspected part by means of discrimination such as an autoencoder, it is not necessary to memorize the individual configuration information of each of the inspected parts in advance, so when inspecting Improve inspection efficiency without referring to this information.

Means 5. The mesh mask inspection device according to any one of the aforementioned means 1 to 4, wherein the aforementioned irradiating means is configured to irradiate three-dimensional measuring light (for example, patterned light having a stripe-shaped light intensity distribution) as The aforementioned light of predestination, The shape data obtaining means is configured to obtain the three-dimensional shape data related to the inspected portion by a predetermined three-dimensional measurement method (for example, a phase shift method).

Here, as an example of the above-mentioned "three-dimensional measurement method", there can be mentioned a phase shift method for obtaining three-dimensional shape data based on a plurality of image data captured under a plurality of patterned lights with different phases.

According to the above method 5, by using the three-dimensional measurement method such as the phase shift method, by obtaining the three-dimensional shape data of the inspected part, it is possible to determine whether the inspected part is abnormal (defective part) with excellent accuracy. As a result, inspection accuracy can be improved.

In addition, in order to properly grasp that during solder printing, cream flux flows from the gap around the opening on the back side of the mesh mask, and is transferred to abnormal parts on the islands that cause penetration or sloshing of the cream flux, Even in the shape data, it is important to obtain the height direction information (height data) of the abnormal part that can generate the aforementioned gap.

In this regard, as shown in this means, by acquiring three-dimensional shape data related to the inspected portion, it is possible to determine with excellent accuracy whether an abnormal portion in which the above-mentioned gap occurs exists in the inspected portion. As a result, the inspection accuracy can be improved.

Means 6. The net mask inspection device according to means 1 to 5, wherein the aforementioned net mask is configured so that the aforementioned back side becomes the lower side, and the aforementioned irradiation means is configured to illuminate the back side of the aforementioned net mask from below The aforementioned predetermined light; and the aforementioned photographing means are configured to photograph the back side of the aforementioned net mask from below.

In the case where inspection is performed from above assuming that the back side (substrate contact surface side) of the mesh mask is placed on the upper side, it may be affected by external light such as indoor lighting, and stable inspection may be difficult.

Regarding this point, according to the above-mentioned means 6, the occurrence of the above-mentioned disadvantages can be suppressed and the inspection accuracy can be improved. In particular, as in the case of using the phase shift method, it is more effective when inspection is performed in an environment that is easily affected by minute luminance changes.

[Mode for Carrying Out the Invention]

One embodiment of the present invention will be described below. First, the structure of the substrate on which solder is printed and components are mounted will be described. FIG. 1 is an enlarged partial plan view of a part of a printed circuit board as a substrate.

As shown in FIG. 1 , a printed circuit board 1 has a copper foil wiring pattern (not shown) or a plurality of islands 3 formed on the surface of a flat base substrate 2 made of glass epoxy resin or the like. Furthermore, a resist film 4 is coated on the surface of the base substrate 2 except for the island portion 3 . Furthermore, as shown in FIGS. 1 and 3 , a viscous cream solder 5 is printed on the island portion 3 . In addition, in FIGS. 1 and 3 , it is expedient to attach scattered dots to the portion where the cream solder 5 is displayed.

Next, a production line (process) for producing the printed circuit board 1 will be described with reference to FIG. 2 . FIG. 2 is a block diagram showing the configuration of a production line 10 for printed substrates 1 . The production line 10 in this embodiment is set to convey the printed circuit board 1 from left to right when viewed from the front side.

As shown in FIG. 2 , the production line 10 is sequentially provided with a solder printing machine 12 , a solder printing inspection device 13 , a component mounting machine 14 , and a reflow device 15 from the upstream side (left side of FIG. 2 ).

The solder printing machine 12 is used to perform a solder printing step of printing cream solder 5 on the island portion 3 of the printed circuit board 1 . As shown in FIG. 3 , the solder printer 12 is provided with: a flat metal mask 20 having a plurality of openings 21 corresponding to the islands 3 on the printed circuit board 1; The scraper 22 that slides on the surface 202. In this embodiment, the metal mask 20 corresponds to the mesh mask of this embodiment.

In addition, the metal mask 20 in this embodiment (refer to FIG. 4 ) is made of a rectangular thin-plate-shaped metal mask body part 20a formed with the aforementioned plurality of openings 21, and a peripheral edge for holding the mask body part 20a. It is composed of rectangular frame-shaped metal frame parts 20b on the four sides. By holding the thin-plate-shaped mask body portion 20a in a tensioned state on the highly rigid frame portion 20b, it is possible to suppress bending of the mask body portion 20a, and the like.

Under the above configuration, in the solder printing step, first, the metal mask 20 is aligned and brought into contact with the surface side of the printed circuit board 1 . Next, the paste solder 5 is supplied to the surface 202 of the metal mask 20 . Then make the scraper 22 slide along the surface 202 of the metal mask 20 , and push the paste solder 5 into the opening 21 .

Then, by detaching the metal mask 20 from the printed circuit board 1, the cream solder 5 filled in the opening 21 is released from the back surface 201 side, and printed (transferred) onto the island 3, the solder printing is completed. Finish.

The solder printing inspection device 13 is used to perform a solder printing inspection step of inspecting the state (for example, printing position, height, amount, etc.) of the cream solder 5 printed as described above.

The component mounting machine 14 performs a component mounting step of mounting an electronic component 25 (see FIG. 1 ) on the island portion 3 on which the cream solder 5 is printed. The electronic component 25 is provided with a plurality of electrodes or lead wires, and each of the electrodes or lead wires is temporarily fixed to predetermined cream solder 5 .

The reflow device 15 is used to heat and melt the cream flux 5, so as to implement the reflow step of soldering (soldering) the electrodes or wires of the island part 3 and the electronic component 25.

In other respects, although not shown, in the production line 10, a conveyor for conveying the printed circuit board 1 is provided between the above-mentioned devices such as the solder printing machine 12 and the solder printing inspection device 13. In addition, a branching device is provided between the solder printing inspection device 13 and the component mounting machine 14 . Furthermore, the printed circuit board 1 judged to be a good product by the solder printing inspection device 13 is guided to the component mounting machine 14 on the downstream side thereof, while the printed circuit board 1 judged to be a defective product is discharged to the defective product storage section by the diverting device. .

Furthermore, a cleaning device 29 and a metal mask inspection device 30 are provided in the vicinity of the solder printer 12 . The metal mask inspection device 30 is configured as a mesh mask inspection device in the present embodiment.

The cleaning device 29 is a cleaning step for cleaning the metal mask 20 to which dirt and the like adhered in the solder printer 12 . In addition, in the present embodiment, in the manufacturing process (line 10) of the above-mentioned printed circuit board 1, every time a predetermined number of times of solder printing is performed by the solder printing machine 12, that is, the solder printing is suspended, and the metal used for the solder printing is used. Mask 20 is cleaned.

The metal mask inspection device 30 is a metal mask 20 that has been cleaned in the cleaning device 29 in the above-mentioned manner, or a new unused metal mask 20 that is to be installed before the solder printer 12, etc. 20 examiners.

Here, the configuration of the metal mask inspection device 30 will be described in detail with reference to FIGS. 4 and 5 . FIG. 4 is a schematic configuration diagram schematically showing a metal mask inspection device 30 . FIG. 5 is a block diagram showing the functional configuration of the metal mask inspection device 30 .

The metal mask inspection device 30 is equipped with: a conveying mechanism 31 for conveying or positioning the metal mask 20; an inspection unit 32 for performing inspection of the metal mask 20; a control device 33 for driving the conveying mechanism 31 or the inspection unit 32 Control is the first, and various controls, image processing, and calculation processing in the metal mask inspection device 30 are executed.

The conveyance mechanism 31 is equipped with: a pair of conveyance rails 31a arranged along the loading and unloading direction of the metal mask 20; an endless conveyer belt 31b arranged rotatably with respect to each conveyance rail 31a; and a motor for driving the conveyance belt 31b. and other driving means (not shown); and a clamping mechanism (not shown) for positioning the metal mask 20 at a predetermined position;

With the above configuration, the metal mask 20 carried into the metal mask inspection device 30 is inserted into the conveying rail 31a by the frame portions 20b at both side edges in the width direction perpendicular to the carrying in and carrying out direction, and placed on the conveyor belt. 31b on. Next, the conveyor belt 31b starts to operate, and conveys the metal mask 20 to a predetermined inspection position. When the metal mask 20 reaches the inspection position, the conveyor belt 31b stops and the clamping mechanism operates. Through the operation of the clamping mechanism, the conveyor belt 31b is pushed up, and the frame portions 20b at both side edges of the metal mask 20 are clamped by the conveyor belt 31b and the upper edge of the conveyor rail 31a. Thereby, the metal mask 20 can be positioned and fixed at the inspection position. When the inspection is finished, the fixation by the clamping mechanism is released, and the conveyor belt 31b starts to move. Thereby, the metal mask 20 is carried out from the metal mask inspection apparatus 30. As shown in FIG. Of course, the structure of the conveyance mechanism 31 is not limited to the above-mentioned form, and other structures can also be employ|adopted.

The inspection unit 32 is arranged below the conveyance rail 31a (conveyance path of the metal mask 20). The inspection unit 32 includes a first illuminating device 32A and a second illuminating device 32B as irradiation means, and irradiates the metal mask 20 with predetermined light rays (pattern light having a stripe-shaped light intensity distribution) for three-dimensional measurement from obliquely below. The predetermined inspection range of the back side 201; the camera 32C, as a means of photographing the predetermined inspection range of the back side 201 of the metal mask 20 from directly below; the X-axis moving mechanism 32D (referring to FIG. 5 ), which can move in the X-axis direction (Fig. 4 left and right direction); and Y-axis moving mechanism 32E (referring to Fig. 5), is set to move toward the Y-axis direction (Fig. 4 front-rear direction);

In addition, in the present embodiment, the back surface 201 on the substrate-contacting surface side of the metal mask 20 that is in contact with the printed circuit board 1 at the time of solder printing is directed downward (becomes the lower surface), and becomes the surface on the substrate non-contacting surface side. The metal mask 20 is set to the metal mask inspection device 30 so that it faces upward (turns to the upper side) at 202 .

Furthermore, the "inspection range" of the rear surface 201 of the metal mask 20 refers to one of the plurality of regions preset in advance on the rear surface 201 of the metal mask 20, taking the size of the imaging angle of view (shooting range) of the camera 32C as a unit. area.

The control device 33 (moving mechanism control part 76) drives and controls the X-axis moving mechanism 32D and the Y-axis moving mechanism 32E, so that the inspection unit 32 can move to any inspection range on the back side 201 of the metal mask 20 positioned and fixed at the inspection position. The lower position moves. Then, by sequentially moving the inspection unit 32 to a plurality of inspection ranges set on the back side 201 of the metal mask 20, and performing inspections related to the inspection ranges, the entire area of the back side 201 of the metal mask 20 can be inspected. Check composition.

The first lighting device 32A includes a first light source 32Aa that emits a predetermined light, or a first liquid crystal shutter that forms a first grid that converts light from the first light source 32Aa into a first pattern light having a stripe-like light intensity distribution. 32Ab, and drive control is performed by the control device 33 (illumination control unit 72 described later).

The second illuminating device 32B includes a second light source 32Ba that emits a predetermined light, or a second liquid crystal shutter 32Bb that forms a second lattice for converting the light from the second light source 32Ba into a second pattern light having a stripe-like light intensity distribution. , and drive control is performed by the control device 33 (illumination control unit 72 described later).

Under the above-mentioned configuration, the light systems emitted from the light sources 32Aa, 32Ba are respectively guided to the condenser lens (not shown in the figure), and after being made into parallel light here, they are guided to the projection lens (not shown in the figure) through the liquid crystal shutters 32Ab, 32Bb. abbreviated), projected onto the metal mask 20 as patterned light. In addition, in this embodiment, switching control of the liquid crystal shutters 32Ab and 32Bb is performed so that the phases of the respective pattern lights are shifted one by one at a quarter pitch.

In addition, by using the liquid crystal shutters 32Ab and 32Bb as grids, it is possible to irradiate pattern light close to an ideal sine wave. According to this configuration, the measurement resolution of the three-dimensional metrology can be improved. In addition, the phase shift control of the pattern light can also be performed electrically, and the simplification of the device can be achieved.

Camera 32C has: CCD (Charge Coupled Device, Charge Coupled Device) type image sensor or CMOS (complementary metal oxide semiconductor, Complementary Metal Oxide Semiconductor) type image sensor and other shooting elements; and the image imaging of metal mask 20 In the optical system (lens unit or aperture, etc.) of the photographing element, its optical axis is arranged along the vertical direction (Z-axis direction). Of course, the imaging elements are not limited to these elements, and other imaging elements may also be used.

The camera 32C is driven and controlled by the control device 33 (camera control unit 73 described later). More specifically, the control device 33 can execute the photographing process by the camera 32C while synchronizing with the illumination process by the two lighting devices 32A and 32B. Thereby, among the light irradiated from the 1st illuminating device 32A or the 2nd illuminating device 32B, the light reflected by the metal mask 20 is captured by the camera 32C, and an image data is generated.

In this way, the image data captured and generated by the camera 32C is converted into a digital signal inside the camera 32C, and can be transferred and stored in the control device 33 (the image acquisition unit 74 described later) in the form of a digital signal. Next, the control device 33 (the data processing unit 75 and the like to be described later) executes various image processing and calculation processing described later based on the image data.

The control device 33 is composed of a CPU (Central Processing Unit, Central Processing Unit) that performs predetermined arithmetic processing, a ROM (Read Only Memory, Read Only Memory) that stores various programs or fixed value data, etc., and performs various arithmetic processing. It consists of RAM (Random Access Memory, Random Access Memory) which temporarily stores various data, and a computer including these peripheral circuits.

Next, the control device 33 is executed by the CPU according to various programs, thereby serving as the main control unit 71, lighting control unit 72, camera control unit 73, image acquisition unit 74, data processing unit 75, moving mechanism control unit 76, Various functional units such as the learning unit 77, the inspection unit 78, and the transport mechanism control unit 79 function.

However, the above-mentioned various functions are realized through the operation of various hardware such as the above-mentioned CPU, ROM, and RAM, and it is not necessary to clearly distinguish functions realized by hardware or software. Some or all of these functions can be implemented by hardware circuits such as ICs. accomplish.

Furthermore, the control device 33 is provided with: an input unit 55 composed of a keyboard, a mouse, a touch panel, etc., a display unit 56 with a display screen function such as a liquid crystal display, and a memory that can store various data or programs, calculation results, etc. Part 57, the communication part 58 etc. which can send and receive various data with the outside.

Hereinafter, the various functional units constituting the control device 33 will be described in detail. The main control unit 71 is a functional unit in charge of the overall control of the metal mask inspection device 30 , and is configured to receive and send various signals from other functional units such as the lighting control unit 72 and the camera control unit 73 .

The lighting control unit 72 is a functional unit for driving and controlling the first lighting device 32A and the second lighting device 32B.

The camera control unit 73 is a functional unit that drives and controls the camera 32C, and controls shooting timing and the like based on command signals from the main control unit 71 .

The video acquisition unit 74 is a functional unit for inputting video data captured by the camera 32C.

The data processing unit 75 is a functional unit that performs predetermined image processing on the image data acquired by the image acquisition unit 74 or performs three-dimensional measurement processing using the image data. For example, in the learning process described later, the shape data for learning (three-dimensional shape data for learning) to be used as learning data for learning of the deep neural network 90 (hereinafter simply referred to as "neural network 90", see Fig. 6 ) is generated. material). In addition, in the inspection process described later, shape data for inspection (three-dimensional shape data for inspection) are generated.

The moving mechanism control unit 76 is a functional unit that drives and controls the X-axis moving mechanism 32D and the Y-axis moving mechanism 32E, and controls the position of the inspection unit 32 according to the command signal from the main control unit 71 .

The learning unit 77 is a functional unit that learns the neural network 90 using learning materials and constructs an AI (Artificial Intelligence) model 100 as a discrimination means.

In addition, the AI model 100 in this embodiment uses the shape data related to the inspected portion EA [see FIG. The generative model constructed by the neural network 90 for deep learning (Deep Learning) has a so-called autoencoder (autoencoder) structure.

The structure of the neural network 90 will now be described with reference to FIG. 6 . FIG. 6 is a schematic diagram conceptually showing the structure of the neural network 90 . As shown in FIG. 6 , the neural network 90 has: an encoder unit 91 as an encoding unit, which extracts a feature quantity (latent variable) TA from the input shape data GA; and a decoder unit 92 as a decoder unit, which extracts the The feature amount TA reconstructs (reconstitutes) the shape data GB to form a structure of a convolutional auto-encoder (CAE: Convolutional Auto-Encoder).

The structure of a convolutional autoencoder is a well-known technology, and its detailed description is omitted, but the encoder unit 91 has a plurality of convolution layers (Convolution Layer) 93, and in each convolution layer 93, a plurality of filters (core part) are used. ) 94 The result of performing the folding operation on the input data is output as the input data of the next layer. Similarly, the decoder unit 92 has a plurality of deconvolution layers (Deconvolution Layer) 95, and in each deconvolution layer 95, the result of performing a deconvolution type operation on the input data using a plurality of filters (core parts) 96 is output as input to the next layer. In addition, in the learning process mentioned later, the weight (parameter) of each filter 94,96 can be updated.

The inspection unit 78 is a functional unit that inspects the presence or absence of an abnormality in the inspected portion EA of the metal mask 20 . In this embodiment, as shown in FIG. 9(a) and (b), the predetermined range of the peripheral portion of one opening 21 on the back side 201 side of the substrate contact side of the metal mask 20 is set to be one Check Ministry EA. That is, a plurality of inspected parts EA can be set on the back surface 201 of the metal mask 20 . FIG. 9( a ) is a partial enlarged plan view of the metal mask 20 showing the opening 21 and its peripheral portion viewed from the surface 202 side. Fig. 9(b) is a sectional view of line A-A of Fig. 9(a). In addition, a scatter pattern is added to the peripheral portion (surface 202) of the opening 21 in FIG. 9( a ), so that the range can be easily distinguished.

The transport mechanism control unit 79 is a functional unit that drives and controls the transport mechanism 31 , and controls the position of the metal mask 20 according to the instruction signal from the main control unit 71 .

The memory unit 57 is composed of HDD (hard disk drive, Hard Disk Drive) or SSD (solid state machine, Solid State Drive), etc., and has, for example, the AI model 100 (the neural network 90 and the learning information obtained by learning it) A predetermined area of memory to be memorized. This memory area constitutes the model storage means in this embodiment.

The communication unit 58 includes, for example, a wireless communication interface conforming to communication standards such as wired LAN (Local Area Network) or wireless LAN, and is configured to transmit and receive various data to and from the outside.

Next, the learning process of the neural network 90 executed by the metal mask inspection device 30 will be described with reference to the flowchart of FIG. 7 .

First, the operator prepares an unused metal mask 20 that has no abnormality (does not have various abnormal parts Q described later). The metal mask 20 prepared here preferably has an opening 21 having the same shape as the metal mask 20 to be inspected. However, the thickness and material of the metal mask 20 , the size and layout of the opening 21 do not have to be identical, so that learners can learn from various types of learning materials, which is better in terms of versatility.

In addition, the operator causes the main control unit 71 to execute a predetermined learning program after arranging the unused metal mask 20 without the aforementioned abnormality at a predetermined inspection position of the metal mask inspection device 30 .

When starting the learning process based on the execution of a predetermined learning program, the main control unit 71 first performs preprocessing for implementing learning of the neural network 90 in step S101.

In this preprocessing, the same processing as the image data acquisition processing (step S301 ) and the three-dimensional shape data acquisition processing (step S302 ) related to the inspection processing described later is performed. In addition, the details of these processes will be described later, so they will be briefly described here.

First, with respect to the predetermined inspection range on the rear surface 201 side of the metal mask 20 , while changing the phase of the first pattern light irradiated from the first illuminating device 32A, imaging processing is performed four times under the first pattern light having different phases. Thereafter, while changing the phase of the second patterned light irradiated from the second illuminating device 32B, imaging processing is performed four times under the second patterned light with different phases, and a total of eight types of area image data are obtained.

Then, on the basis of the 4 sets of area image data captured under each of the above-mentioned patterned lights, the three-dimensional shape measurement of the predetermined inspection range including a plurality of inspected parts EA (openings 21) is carried out by using the known phase shift method , and the measurement result (area shape data) is stored in the memory unit 57 .

In addition, the above-mentioned series of processing is performed on the side of the metal mask 20 before obtaining the area shape data including the shape data related to the required number of inspected parts EA (peripheral parts of the opening 21) as learning data. Move within the inspection range, while doing it repeatedly.

In step S101, when the area shape data including the shape data related to the examined part EA required for learning is obtained, in the next step S102, according to the instruction from the main control part 71, the learning part 77 will prepare Learning Neural Networks 90. For example, the neural network 90 previously stored in the memory unit 57 or the like is read out. Alternatively, the neural network 90 is constructed based on network configuration information (for example, the number of layers of the neural network, the number of nodes in each layer, etc.) to be stored in the memory unit 57 or the like.

In step S103, shape data for learning (three-dimensional shape data for learning) are acquired as learning data. Specifically, according to the instruction from the main control unit 71, the data processing unit 75 will extract the area shape data from the plurality of inspected parts EA included in the area shape data based on the area shape data stored in the memory unit 57 in step S101. One inspected part EA is used, and the shape data related to the inspected part EA is acquired as one shape data for learning. Then, the shape data for learning is output to the learning unit 77 . That is, only unused shape data related to the inspected portion EA of the metal mask 20 without abnormality is used as learning data (shape data for learning).

In step S104, the reconstructed shape data is obtained. Specifically, according to the instruction from the main control unit 71, the learning unit 77 will send the shape data for learning obtained in step S103 as input data to the input layer of the neural network 90, thereby obtaining the shape data from the neural network 90. The reconstructed shape data output by the output layer.

In the following step S105, the learning part 77 will compare the shape data for learning obtained in step S103 with the reconstructed shape data output by the neural network 90 in step S104, and determine whether the error is sufficiently small (whether it is within below a predetermined threshold).

Here, if the error is sufficiently small, the neural network 90 and its learning information (updated parameters described later, etc.) are stored in the memory unit 57 as the AI model 100, and this learning process ends.

On the other hand, if the error is not small enough, in step S106, the network update process (learning of the neural network 90) is performed, and then returns to step S103 to repeat the series of processes described above.

Specifically, in the network update process of step S105, known learning algorithms such as error backpropagation (Backpropagation) are used, in order to set the weights (parameters) of the above-mentioned filters 94, 96 in the neural network 90 to ) is updated to a more appropriate one so that the loss function representing the difference between the learning shape data and the reconstructed shape data is minimized. In addition, for example, BCE (Binary Cross-entropy) may also be used as the loss function.

By repeating these processes several times, in the neural network 90, the error between the shape data for learning and the reconstructed shape data becomes extremely small, and more accurate reconstructed shape data can be output.

Next, the metal mask inspection process executed by the metal mask inspection device 30 will be described with reference to the flowchart of FIG. 8 . However, the inspection process shown in FIG. 8 is performed for each predetermined inspection range of the metal mask 20 .

When the metal mask inspection device 30 is carried into the metal mask 20 and positioned at a predetermined inspection position, the inspection process starts according to the execution of a predetermined inspection program.

When the inspection process is started, first, the image data acquisition process is executed in step S301. In this embodiment, in the inspection related to each inspection range on the back surface 201 side of the metal mask 20, the phase of the first patterned light irradiated from the first illuminating device 32A is changed, and the first patterned light with a different phase After executing the photographing process four times, while changing the phase of the second pattern light irradiated from the second illumination device 32B, the photographing process is performed four times under the second pattern light with different phases, and a total of 8 sets of images are obtained. material. The details are as follows.

As described above, when the metal mask 20 carried into the metal mask inspection device 30 is positioned and fixed at a predetermined inspection position, according to the instruction from the main control unit 71, the moving mechanism control unit 76 first drives and controls the X-axis moving mechanism 32D and The Y-axis moving mechanism 32E moves the inspection unit 32 to align the shooting angle (shooting range) of the camera 32C with the predetermined inspection range of the back surface 201 of the metal mask 20 .

In addition, the lighting control unit 72 switches the liquid crystal shutters 32Ab and 32Bb for controlling the two lighting devices 32A and 32B, and sets the positions of the first grid and the second grid formed on the two liquid crystal shutters 32A and 32Bb at predetermined reference positions. .

When the switching setting of the first grid and the second grid is completed, the lighting control unit 72 lights up the first light source 32Aa of the first lighting device 32A to irradiate the first pattern light, and the camera control unit 73 drives and controls the camera 32C. , execute the first shooting process under the first pattern light. In addition, the video data generated by the photographing process is taken into the video acquisition unit 74 at any time (the same applies hereinafter). According to such a configuration, the area image data of the inspection range including a plurality of inspected parts EA (openings 21 ) can be obtained.

Then, the lighting control unit 72 turns off the first light source 32Aa of the first lighting device 32A and executes switching processing of the first liquid crystal shutter 32Ab at the same time as the first imaging process under the first pattern light is completed. Specifically, the position of the first lattice forming the first liquid crystal shutter 32Ab is switched from the reference position to the second position where the phase of the first pattern light is shifted by 1/4 pitch (90°).

When the switch setting of the first grid is completed, the lighting control unit 72 will light the light source 32Aa of the first lighting device 32A to illuminate the first pattern light. Next, the second shooting process is executed. Then, by repeatedly performing the same process, four sets of image data of the region under the first pattern light with a phase difference of 90° successively are obtained.

Next, the lighting control unit 72 turns on the second light source 32Ba of the second lighting device 32B to irradiate the second pattern light, and drives and controls the camera 32C with the camera control unit 73 to carry out the first lighting operation under the second pattern light. shooting processing.

Then, when the first imaging process under the second pattern light ends, the illumination control unit 72 turns off the second light source 32Ba of the second illumination device 32B, and executes the switching process of the second liquid crystal shutter 32Bb. Specifically, the position of the second lattice formed on the second liquid crystal shutter 32Bb is switched from the aforementioned reference position to the second position where the phase of the second pattern light is shifted by 1/4 pitch (90°). .

When the switching setting of the second grid is completed, the lighting control part 72 will light the light source 32Ba of the second lighting device 32B, and illuminate the second pattern light. 2 Processing of the second shot under patterned light. Thereafter, by repeating the same process, four kinds of area image data under the second pattern light whose phases successively differ by 90° are obtained.

In the next step S302, the acquisition process of the three-dimensional shape data is performed. Specifically, according to the instruction from the main control unit 71, the data processing unit 75 will, in the above step S301, carry out a process including a plurality of subject areas through a known phase shift method based on the four kinds of area image data captured under each pattern light in the above-mentioned step S301. The three-dimensional shape of the predetermined inspection range of the inspection part EA (opening part 21 ) is measured, and the related measurement results (area shape data) are stored in the memory part 57 . The shape data acquisition means of this embodiment is constituted by the function of executing related processing. In addition, in the present embodiment, since the three-dimensional shape is measured by irradiating the patterned light from two directions, it is possible to prevent the generation of the shadow portion where the patterned light is not irradiated.

Next, in step S303, shape data for inspection (three-dimensional shape data for inspection) related to each inspected portion EA is acquired.

Specifically, first, according to the instruction from the main control unit 71, the data processing unit 75 will specify and retrieve all the complex numbers contained in the area shape data according to the area shape data related to the predetermined inspection range obtained in the above step S302. The inspected department EA.

Then, these are individually numbered and registered as original shape data related to the inspected portion EA. In this process, for example, the original shape data related to the non-abnormal inspected part EA as shown in Fig. 10(a), or the inspected part EA with some abnormalities as shown in Fig. 10(b) to (e) is acquired. The original shape data involved, etc.

In step S304, reconstruction processing (reconstruction shape data acquisition processing) is performed. The reconstructed shape data acquisition means in this embodiment is constituted by the function of performing this processing.

Specifically, according to an instruction from the main control unit 71 , the inspection unit 78 inputs the original shape data related to the inspected part EA of a predetermined number (for example, No. 001) obtained in step S303 into the input layer of the AI model 100 . Next, the shape data reconstructed by the AI model 100 and output from the output layer is obtained as the reconstructed shape data related to the inspected part EA of the aforementioned predetermined number (for example, No. 001), and at the same time, the reconstructed shape data and the same number The original shape data is associated and memorized. In this manner, in this process, reconstructed shape data can be acquired for all the parts to be inspected EA whose numbers are registered in step S303.

Here, it is natural for the AI model 100 to input the original shape data related to the inspected part EA without abnormality as shown in FIG. In the case of the input of the original shape data related to the abnormal inspected part EA as shown, by learning in the above-mentioned manner, the shape data related to the non-abnormal inspected part EA as shown in Figure 10 (a) can be output as a reconstructed shape data.

In step S305, a judgment process of good or bad is executed. Specifically, according to the command from the main control unit 71 , the checking unit 78 first compares the original shape data and the reconstructed shape data of the same number, and extracts the difference between the two shape data. The comparison means in this embodiment is constituted by the function of performing such processing.

Next, the inspection unit 78 determines whether the difference between the two shape data, that is, the part corresponding to the abnormal part Q is greater than a predetermined threshold. Here, when the difference between the two image data is larger than a predetermined critical value, it is determined as "abnormal".

For example, as shown in FIGS. 9( a ) and ( b ), when there is an abnormal part Q in the inspection part EA on the back surface 201 side of the metal mask 20 , the difference between the depth d and the width w of the abnormal part Q When each judgment item exceeds a predetermined threshold value, it is judged to be "abnormal".

According to this configuration, for example, as shown in FIG. 10( b ), when the chamfered portion Q1 around the opening 21 on the back surface 201 side of the metal mask 20 becomes too round due to excessive electrolytic polishing, the chamfered The corner Q1 is also regarded as an abnormal part Q.

Similarly, as shown in FIG. 10( c ), when there is a wound Q2 such as a depression or a notch connected to the opening 21 in the inspected portion EA, or as shown in FIG. 10( d ), the periphery of the opening 21 is deformed. In case of Q3, it is regarded as an abnormal part Q according to its degree.

On the other hand, when the difference between the two image data is smaller than a predetermined threshold value, it is determined as "no abnormality". The judging means in this embodiment is constituted by performing these judging functions.

Here, when the inspection unit 78 determines that all the inspected parts EA included in the area shape data (planned inspection range of the metal mask 20) are "no abnormality", the inspection range related to the area shape data It is judged as "good", and the result is stored in the storage unit 57, and this process ends.

On the other hand, if there is only one inspected part EA judged to be "abnormal" among the plurality of inspected parts EA included in the area shape data (planned inspection range of the metal mask 20), the The inspection range related to the area shape data is judged to be "defective", the result is stored in the storage unit 57, and this process ends.

Next, the metal mask inspection device 30 executes the result of the above-mentioned inspection process for all the inspection ranges of the back surface 201 of the metal mask 20, and when all the inspection ranges are judged to be "good", that is, it is judged that there is no abnormal metal mask 20 (determination pass), and inform the operator of its intention through the display unit 56 or the communication unit 58, etc.

On the other hand, when there is one inspection range judged as "defective" among all the inspection ranges on the metal mask 20, the metal mask inspection device 30 judges (fail judgment) that there is an abnormal metal mask 20. , and notify the operator of this fact via the display unit 56 or the communication unit 58 .

As described in detail above, in this embodiment, of the front and back surfaces of the metal mask 20, the back surface 201 side that will be the substrate contact surface side that contacts the printed circuit board 1 during solder printing is implemented and opened. The inspection related to the peripheral part of the part 21 , that is, the planned part to be inspected EA.

Thereby, various abnormal parts Q (Q1 to Q4) related to the periphery of the opening 21 (the part to be inspected EA) on the back side 201 side of the metal mask 20, which cause penetration or sloshing of the cream solder 5 during printing, etc. can be detected. As a result, occurrence of poor printing can be suppressed.

In particular, in this embodiment, the AI model 100 constructed using the learning neural network 90 is used to determine whether or not there is an abnormal part Q ( Q1 to Q4 ) in the inspected part EA of the metal mask 20 . According to this constitution, minute abnormalities which were difficult to detect in the conventional art can be detected.

Furthermore, in this embodiment, because the original shape data obtained by photographing the inspected part EA will be compared with the reconstructed shape data reconstructed based on the original shape data, there will be no difference between the two compared shape data. Shooting conditions on the side of the metal mask 20 belonging to the inspection object (for example, the arrangement position or arrangement angle of the metal mask 20, bending, etc.), or shooting conditions on the side of the metal mask inspection device 30 (for example, lighting conditions or camera 32C Due to the influence of the difference in the picture angle, etc.), it is possible to more accurately detect the finer abnormal part Q.

In addition, when it is assumed that an inspection is performed on a predetermined inspected portion EA (peripheral portion of the opening 21 ) at a predetermined position of the metal mask 20 , configuration information (position data, dimension data, etc.) of the inspected portion EA is required. In the configuration as the standard of good/failure judgment, board design information such as Gerber data is stored in advance, and the configuration information of the inspected part EA to be inspected is obtained appropriately, and compared with the configuration information, the process to become Because of the good/failure judgment of the inspected portion EA, which is the inspection object, there is a possibility that the inspection efficiency may decrease. Moreover, the positioning of the metal mask 20 to the inspection position must also be performed correctly.

In this regard, according to the present embodiment, since the AI model 100 learned for the inspected part EA is used to perform the inspection of each inspected part EA, it is not necessary to store the configuration information of each of the plurality of inspected parts EA in advance. , so it is not necessary to refer to these materials during the inspection, so the inspection efficiency can be improved.

In addition, in the present embodiment, during solder printing, the back surface 201 of the substrate contact surface side of the metal mask 20 to be in contact with the printed circuit board 1 is arranged downward, and the solder is printed from the bottom of the metal mask 20 . The inspection is performed by the inspection unit 32 .

According to this configuration, it is less likely to be affected by external disturbance light such as indoor lighting, and the inspection accuracy can be improved. In particular, it is more effective when inspection can be performed in an environment that is not easily affected by subtle brightness changes, such as when using the phase shift method.

In addition, it is not limited to the description content of the said embodiment, For example, it can implement as follows. Of course, other application examples and modification examples not exemplified below are also possible.

(a) The configuration of the mesh mask is not limited to the metal mask 20 according to the above-mentioned embodiment, and other configurations may be employed.

(a-1) For example, the metal mask 20 according to the above embodiment is composed of the mask body 20a and the frame 20b, but the present invention is not limited thereto, and the frame 20b may be omitted.

(a-2) The mask body part 20a according to the above embodiment is formed of a metal material. However, it is not limited to this, and as the mask body part 20a, for example, a coating part formed by coating a photosensitive emulsion or the like on a "yarn (mesh)" made of nylon, polyester, stainless steel, etc. wire rods may be used. , and only the part corresponding to the opening 21 exposes the "yarn" type; a composite product formed by pasting a metal plate with the opening 21 on the "yarn (screen)".

(b) The configuration related to the production of the printed circuit board 1 such as the production line 10 is not limited to the above-described embodiment, and other configurations may be employed.

(b-1) In the above embodiment, the cleaning device 29 and the metal mask inspection device 30 are installed in the vicinity of the solder printing machine 12 together. However, the present invention is not limited thereto, and the functions of the cleaning device and/or the metal mask inspection device may be integrated with the solder printing machine 12 or the solder printing inspection device 13 .

(b-2) The cleaning device 29 or the metal mask inspection device 30 may be independently installed in other spaces such as the production line 10 .

(c) The configuration related to the metal mask inspection device 30 is not limited to the above-mentioned embodiment, and other configurations may be employed.

For example, in the above-mentioned embodiment, it may be arranged such that the back surface 201 of the metal mask 20 that is in contact with the printed circuit board 1 faces downward during solder printing, and the metal mask 20 may be inspected by the inspection unit 32. The lower side implements the composition of the inspection.

It is not limited to this, and it may be arranged so that the back surface 201 (substrate abutting surface) side of the metal mask 20 faces upward, and the inspection unit 32 arranged above the metal mask 20 may perform inspection from the upper side. . However, in order to improve the inspection accuracy by suppressing the influence of external light such as indoor lighting, it is preferable to perform the inspection with the rear surface 201 side of the metal mask 20 facing downward.

(d) The configuration of the AI model 100 (neural network 90 ) as a discrimination means and its learning method are not limited to the above-mentioned embodiments.

(d-1) In the above-mentioned embodiment, especially (not mentioned above) when performing the learning process of the neural network 90 and the reconstruction process in the metal mask inspection process, it is also possible to execute various data as needed. Formation of processing such as normalization.

(d-2) The structure of the neural network 90 is not limited to that shown in FIG. 6 , and for example, a pooling layer may be provided after the convolutional layer 93 . Of course, the number of layers of the neural network 90, the number of nodes in each layer, the connection structure of each node, etc. may also be configured differently.

(d-3) In the above embodiment, although the AI model 100 (neural network 90) is a generative model having a structure of a convolutional autoencoder (CAE), it may be, for example, a variational autoencoder (VAE : Variational Autoencoder) and other generative models with different forms of autoencoder construction.

(d-4) In the above embodiment, the neural network 90 learns by the error backpropagation method, but it is not limited to this, and it is also possible to make a learning structure using other various learning algorithms.

(d-5) The neural network 90 may also be configured by a circuit dedicated to AI processing such as a so-called AI chip. In this case, only learning information such as parameters is memorized in the memory unit 57 , and it is read out and set in the neural network 90 by a circuit dedicated to AI processing to form the AI model 100 .

(d-6) In the above-mentioned embodiment, a learning unit 77 may be provided to execute the learning of the neural network 90 in the control device 33, but it is not limited thereto. The neural network 90) may be memorized in the memory unit 57, or the learning unit 77 may be omitted. Therefore, learning of the neural network 90 may be performed outside the control device 33 and stored in the memory unit 57 .

(e) The inspection method of the metal mask 20 (inspected part EA) is not limited to the above-mentioned embodiment, and other configurations may be employed.

(e-1) For example, in the above embodiment, it may be configured to determine whether there is an abnormal part Q in the inspected part EA by using the AI model (generated model) 100 to extract the difference between the original shape data and the reconstructed shape data.

It is not limited thereto, and the AI model 100 may not be used, for example, the pre-stored reference shape data (configuration information of the inspected part EA such as position data or size data) is compared with the calculated shape data of the inspected part EA, To determine whether there is an abnormal part Q in the inspected part EA.

(e-2) In the above-mentioned embodiment, when there is an abnormal part Q in the inspected part EA, when the two judgment items such as the depth d and the width w of the abnormal part Q exceed predetermined critical values, namely It was judged as "abnormal".

It is not limited thereto, for example, it is configured such that when at least one of the determination items of the depth d and the width w of the abnormal portion Q exceeds a predetermined threshold value, it is determined that “there is an abnormality”.

Furthermore, it is configured to add more judgment items, and when at least one judgment item or a predetermined plurality of judgment items exceeds a predetermined threshold value among the depth d, width w, length l, and volume v of the abnormal part Q, it is judged to be "There are exceptions." Of course, only one determination item may be set.

(e-3) The type of abnormal part Q to be detected is not limited to those exemplified in the above embodiment, and a configuration capable of detecting other abnormal parts Q may be employed. For example, in the above-mentioned embodiment, as the abnormal part Q, the chamfered part Q1 [refer to FIG. c)], deformation Q3 of the periphery of the opening 21 [see FIG. 10(d)], and the like.

It is not limited thereto, and may be configured, for example, as shown in FIG.

In addition, there is very little possibility of foreign matter Q4 adhering to the unused metal mask 20. For example, when the metal mask 20 is cleaned by the cleaning device 29 and the metal mask 20 is used again through the solder printer 12. In the peripheral part of the opening 21 (the inspected part EA) on the back side 201 side of the metal mask 20, there will be small residues such as cream solder 5 or flux (flux) that cannot be completely removed by cleaning means. The situation remaining in the state.

In this way, when the foreign matter Q4 adheres to the part to be inspected EA, the part to be inspected EA (peripheral portion of the opening 21 on the back surface 201 side of the metal mask 20) floats up when printing flux, which may cause problems. Potential causes of paste flux flow, penetration or sloshing, etc.

(e-4) In the above-mentioned embodiment, when the metal mask 20 (inspected part EA) is inspected, three-dimensional measurement is performed by irradiating patterned light, but instead, it may be configured by irradiating uniform light on one surface. The abnormal part Q is detected by taking the two-dimensional luminance image data related to the inspected part EA obtained by shooting.

In this case, for example, using another AI model (a learned neural network) different from the above-mentioned AI model 100, based on the above-mentioned two-dimensional luminance image data, it may be configured such that the uneven shape of the part EA to be inspected, etc. The difference in reflectance (brightness value of reflected light), etc., and the shape data related to the inspected part EA are obtained.

(f) The three-dimensional measurement method is not limited to the above-mentioned embodiment, and other configurations may also be adopted.

(f-1) For example, in the above embodiment, on the basis of performing three-dimensional measurement by the phase shift method, four types of image data in which the phases of each pattern light are successively different by 90° are obtained, but the number of phase shifts and The amount of phase shift is not limited to this. Other phase shift times and phase shift amounts that can be measured three-dimensionally by the phase shift method can also be used.

For example, it can also be configured to obtain three types of image data with successive phase differences of 120° (or 90°) for three-dimensional measurement, or to obtain two types of image data with successive phase differences of 180° (or 90°). Composition for 3D measurement.

(f-2) In the above-mentioned embodiment, although the phase shift method is used as the three-dimensional measurement method, it is not limited to this, and other three-dimensional measurement methods such as the light cutting method, the moire method, focusing, and the spatial code method can also be used. Law.

1: Printed substrate 3: Island Department 5: Paste flux 10: Production line 12: Solder printing machine 20: metal mask 21: opening 29: Cleaning device 30: Metal mask inspection device 32: Check unit 32A: 1st lighting device 32B: Second lighting device 32C: camera 33: Control device 77: Learning Department 78: Inspection Department 90: Neural Networks 100: AI model 201: back 202: surface EA: inspected department Q: Abnormal Department

FIG. 1 is an enlarged partial top view of a part of a printed substrate. Fig. 2 is a block diagram showing the configuration of a production line for printed circuit boards. FIG. 3 is a schematic diagram for explaining a printing operation by a solder printer. FIG. 4 is a schematic configuration diagram schematically showing a metal mask inspection device. Fig. 5 is a block diagram showing the functional configuration of the metal mask inspection device. FIG. 6 is a schematic diagram illustrating the structure of a neural network. FIG. 7 is a flow chart of the learning process of the neural network. FIG. 8 is a flowchart showing the flow of inspection processing. In FIG. 9 , (a) is a partial enlarged plan view of the metal mask showing the opening and its peripheral portion viewed from the surface side, and (b) is a cross-sectional view along line AA of (a). In Fig. 10, (a) is a partially enlarged cross-sectional view of the metal mask showing the opening without abnormalities and its surrounding parts, and (b) to (e) are the metal masks showing the abnormal opening and its surrounding parts A partial enlarged cross-sectional view.

20: metal mask

20a: mask body part

20b: mask frame part

21: opening

30: Metal mask inspection device

31: Transport mechanism

31a: Transport track

31b: conveyor belt

32: Check unit

32A: 1st lighting device

32Aa: 1st light source

32B: Second lighting device

32Ba: Second light source

32Ab, 32Bb: liquid crystal shutter

32C: camera

201: back

202: surface

Claims (5)

A mesh mask inspection device having a plurality of openings for inspecting a mesh mask used for solder printing on a substrate, characterized by having at least one irradiating means capable of irradiating predetermined light on the A portion to be inspected, which is the peripheral portion of the opening on the back side of the substrate abutting surface side that will be in contact with the substrate during solder printing, among the front and back surfaces of the mesh mask; photographing The method is to photograph the part to be inspected that is irradiated by the predetermined light; the means to obtain shape data is to obtain the image data related to the part to be inspected based on the image data related to the part to be inspected that is photographed by the aforementioned shooting means. The shape data of the aforementioned shape data; and the judging means, based on the shape data related to the aforementioned inspected part obtained by the aforementioned shape data acquisition means, to judge whether the aforementioned inspected part is good or not, and has a discrimination means, which is based on the input from the The neural network of the encoding unit that extracts feature data from shape data and the decoding unit that reconstructs shape data from the feature data uses only the shape data related to the inspected part without abnormalities as learning data, regardless of the layout. And generate, the above-mentioned judging means has: reconstruction shape data obtaining means, can obtain and the above-mentioned inspected part The reconstructed shape data related to the aforementioned inspected part is based on the shape data related to the aforementioned inspected part obtained by the aforementioned shape data acquisition means as the original shape data, and input into the aforementioned identification means for reconstruction and a comparison means, which can compare the aforementioned original shape data with the aforementioned reconstructed shape data; and is configured to judge whether the aforementioned inspected parts are good or not according to the comparison results obtained by the aforementioned comparison means for each of the aforementioned inspected parts. The net mask inspection device according to claim 1, wherein the judging means is configured to judge whether the part to be inspected is good or not by judging whether the size of the chamfer at the periphery of the opening exceeds a predetermined amount. Such as the net mask inspection device of claim 1, wherein the aforementioned irradiation means is configured to be capable of irradiating three-dimensional measurement light as the aforementioned predetermined light; the aforementioned shape data acquisition means is configured to obtain and obtain the predetermined three-dimensional measurement method. Three-dimensional shape data related to the part to be inspected. The net mask inspection device according to claim 1, wherein the judging means is configured to judge whether the inspected part is good or not by judging whether there is an abnormal part exceeding a predetermined size in the inspected part. The mesh mask inspection device according to any one of claims 1 to 4, wherein the mesh mask is arranged so that the back side becomes the lower side, and the irradiation means is arranged so as to be able to illuminate the back side of the mesh mask from The predetermined light is irradiated from below, and the photographing means is arranged so that the rear side of the mesh mask can be photographed from below.

Images