Printing result prediction device and printing system

Abstract

The objective of this invention is to provide a printing result prediction apparatus and printing system capable of improving the accuracy of predicting the volume of solder during printing. The printing result prediction apparatus of an embodiment predicts the volume of solder printed by a printer that prints solder on a substrate. The computation unit of the printing result prediction apparatus inputs printing conditions into a prediction model pre-trained by machine learning, and predicts the volume of solder to be printed during the next printing under the input printing conditions. The printing conditions include mask information, solder information, substrate information, and apparatus parameters. The mask information relates to the mask used in printing. The solder information relates to the solder used in printing. The substrate information includes unevenness information related to the protrusions provided on the substrate used in printing. The apparatus parameters indicate the operation of the printer during printing.

Description

Technical Field

[0001] The present invention relates to a printing result prediction device and a printing system. Prior Technology

[0002] In the manufacturing of mounting substrates, there are: a printing process, which uses a printer to print solder on the substrate; a mounting process, which uses a surface mount machine to mount components on the substrate; and a reflow process, which uses a reflow oven to heat and bond the components to the substrate. The operation of the printing press in the printing process is specified by the printing press's device parameters. The volume of solder printed on the substrate varies depending on the printing conditions, such as the printing press's device parameters. In such a printing process, defects such as excessive or insufficient volume of printed solder sometimes occur.

[0003] [Previous Technical Documents] [Patent Literature] [Patent Document 1] Japanese Patent Application Publication No. 2022-076323 Summary of the Invention

[0004] [The problem that the invention aims to solve] The problem that the invention aims to solve is to provide a printing result prediction device and printing system that can improve the accuracy of predicting the volume of solder in printing.

[0005] [Technical means to solve the problem] An implementation of a printing result prediction device predicts the volume of solder to be printed by a printing press, which prints solder on a substrate by filling an opening in a mask disposed on the substrate. The printing result prediction device includes a computation unit. The computation unit inputs printing conditions into a prediction model pre-trained by machine learning, and predicts the volume of solder to be printed through the opening in the mask during the next printing under the input printing conditions. The printing conditions include mask information, solder information, substrate information, and device parameters. The mask information relates to the mask used in printing. The solder information relates to the solder used in printing. The substrate information includes unevenness information related to the protrusions provided on the substrate used in printing. The device parameters indicate the operation of the printing press during printing. Simple Explanation of the Diagram

[0006] Figure 1 is a block diagram illustrating a printing system including a printing result prediction device in an embodiment. Figure 2(a) and Figure 2(b) are schematic plan views illustrating the substrate and mask used in the printing process. Figures 3(a) to 3(c) are schematic cross-sectional views illustrating the printing process performed by the printing press. Figure 4 is a schematic diagram illustrating the prediction process performed by the printing result prediction device in the implementation mode. Figures 5(a) to 5(h) are schematic cross-sectional views illustrating the printing process. Figures 6(a) to 6(h) are schematic cross-sectional views illustrating the printing process. Figure 7 is a schematic diagram illustrating the image entropy containing silkscreen information. Figure 8 is a schematic cross-sectional view illustrating the printing process performed by a printing press. Figure 9 is a bar chart illustrating the accuracy of the predicted volume of printed solder. Figure 10 is a schematic diagram illustrating a variation in the prediction made by the printing result prediction device in the implementation mode. Figure 11 is a schematic diagram illustrating the prediction made by the result prediction device in the implementation mode. Figure 12 is a schematic plan view illustrating a portion of the substrate. Figure 13(a) and Figure 13(b) are schematic plan views illustrating a portion of the substrate, and Figure 13(c) is a curve diagram illustrating the unevenness of the substrate. Figure 14 is a schematic diagram illustrating the substrate and mask in the printing process. Figure 15 is a graph illustrating the relationship between the distance between the protrusion and the pad and the rate of increase in solder volume. Figure 16 is a schematic diagram illustrating another example of the prediction performed by the result prediction device in the implementation mode. Figure 17 is a bar chart illustrating the prediction error of the volume of printed solder. Figure 18 is a schematic plan view illustrating a portion of the substrate. Figure 19 is a schematic diagram illustrating the learning process of the prediction model. Figure 20 is a schematic diagram illustrating the operation of the printing system in its implementation form. Figure 21 is a schematic diagram illustrating the configuration of a printing result prediction device in an exemplary embodiment. Implementation

[0007] Hereinafter, various embodiments of the present invention will be described with reference to the drawings. In the specification and figures of this application, the same symbols are used for elements that are the same as those already described, and detailed descriptions are appropriately omitted.

[0008] Figure 1 is a block diagram illustrating a printing system including a printing result prediction device in an embodiment. As shown in Figure 1, the printing system 200 of the embodiment includes a printing result prediction device 100, a printing press 110, and an inspection machine 120.

[0009] Printing machine 110 performs the printing process of soldering onto a substrate. Printing machine 110 repeats the printing process. That is, printing machine 110 sequentially performs the printing process on a plurality of substrates, printing solder onto each of the plurality of substrates. Inspection machine 120 inspects not only the shape or amount (volume) of the solder printed onto the substrate by printing machine 110, but also the height and any defects.

[0010] The printing result prediction device 100 is an information processing device (information processing system) that predicts the volume of solder printed on the substrate by the printing process of the printing machine 110.

[0011] The print result prediction device 100 includes an acquisition unit 10 and a processing unit 11. The acquisition unit 10 acquires information from an external source. For example, the acquisition unit 10 may include a communication module, communication interface, or connection terminal for communicating with external machines. For instance, the acquisition unit 10 can be communicatively connected to external machines such as a printing press 110 or an inspection machine 120 to receive information from the external machine. Communication can be performed using either wired or wireless methods. Furthermore, the acquisition unit 10 may include an input interface (keyboard or touch panel, etc.) for a user to input information into the print result prediction device 100. The acquisition unit 10 accepts user input. In this way, the acquisition unit 10 acquires various types of information. The processing unit 11 can communicate with the acquisition unit 10 and can acquire the information acquired by the acquisition unit 10.

[0012] The arithmetic unit 11 or the acquisition unit 10 can be communicatively connected to the memory unit 13 (memory device). The memory unit 13 stores, for example, information about the substrate, information about the mask, and a prediction model used in predicting the volume of solder during printing. The arithmetic unit 11 can acquire the information stored in the memory unit 13. The memory unit 13 can be part of the printing result prediction device 100.

[0013] The calculation unit 11 calculates an estimated volume of solder printed on the substrate by the printing process, for example, based on information obtained from at least one of the printer 110, the inspection machine 120, and the memory unit 13. Specifically, the calculation unit 11 inputs printing conditions into the prediction model and predicts the volume of solder to be printed when the next printing process is performed under the input printing conditions.

[0014] Furthermore, printing conditions are a combination of various parameters that represent the characteristics of a printing process. The printing process is performed under the input printing conditions; in other words, a printing process exhibiting the characteristics indicated by the parameters of those printing conditions is performed. Details regarding the input printing conditions will be described later with reference to Figure 4.

[0015] Referring to Figures 2(a) to 3(c), the printing process performed by the printing press 110 will be described. Figure 2(a) and Figure 2(b) are schematic plan views illustrating the substrate and mask used in the printing process. Figure 2(a) illustrates a portion of the substrate S for solder printing by the printer 110. Figure 2(b) illustrates a portion of the mask M used in printing solder onto the substrate S.

[0016] As shown in Figure 2(a), the substrate S has a substrate B and pads P (conductive portions). The substrate B is, for example, an insulating layer containing an insulator such as resin. A wiring formed of metal is disposed within the substrate S. The pads P are electrically connected to the wiring disposed within the substrate S and are exposed on the surface of the substrate S. The pads P are, for example, copper foil. A plurality of pads P are disposed on one substrate S.

[0017] A screen print SK can be printed on the surface of the substrate S as needed. The screen print SK is, for example, ink printed on the substrate, and has the shape of text, symbols, or graphics. The screen print SK may display information about the substrate S or information related to the components mounted on the substrate S (such as model number or component orientation).

[0018] As shown in Figure 2(b), the mask M is, for example, a plate-shaped metal mask with a plurality of openings H (holes). The position and shape of each opening H correspond to the position and shape of each pad P on the substrate S. In the printing process, solder is applied to each opening H by filling each opening H with solder, so that the solder adheres to each pad P corresponding to each opening H.

[0019] Figures 3(a) to 3(c) are schematic cross-sectional views illustrating the printing process performed by the printing press. As shown in Figure 3(a), a mask M corresponding to the substrate S is disposed on the substrate S to be printed. This covers the surface of the substrate S with the mask M. At this time, a pad P corresponding to the opening H is exposed at the opening H of the mask M.

[0020] A paste-like solder 20 is applied onto the mask M. A scraper 25 is moved along the upper surface of the mask M while in contact with it. In this way, the solder 20 on the mask M is spread and coated onto the mask M by the scraper 25.

[0021] As shown in Figure 3(b), solder 20 is applied to the mask M, and the solder 20 fills each opening H of the mask M.

[0022] As shown in Figure 3(c), the mask M is removed from the substrate S. Thereby, the solder 20 filling the openings H of the mask M is transferred to the pads P of the substrate S. In this way, solder 20 is printed on the substrate S. That is, a layer of solder 20 with a shape corresponding to the openings H is formed.

[0023] Printing machine 110, for example, uses a mask M to sequentially perform such printing processes on a plurality of substrates S. In this way, printing machine 110 sequentially prints solder on a plurality of substrates S. Mechanisms for conveying substrates or masks, moving squeegees, and applying solder can be implemented using appropriate drive devices such as actuators with motors. For example, the control circuit of the printing press 110 controls the operation of the drive device, causing the drive device to perform actions specified by the device parameters.

[0024] Inspection machine 120 detects the volume of each solder 20 printed on each pad P through each opening H. Inspection machine 120, for example, optically measures the shape (height or width) of each solder 20 on each pad P and calculates the volume of each solder 20 on each pad P. Inspection machine 120, for example, shines light on the solder 20 and measures the reflected light. Not limited to this, inspection machine 120 can also detect the volume of the solder 20 by any other method.

[0025] Figure 4 is a schematic diagram illustrating the prediction process performed by the printing result prediction device in the implementation mode. For example, the printing press 110 performs a printing process in which solder is printed on the pad P1 of the substrate S1 by filling the opening H1 of the mask M1 disposed on the substrate S1 with solder. Furthermore, the substrate S1, the mask M1, the opening H1, and the pad P1 are examples of the aforementioned substrate S, mask M, opening H, and pad P.

[0026] As described above, the printing press 110 repeats the printing process. The inspection machine 120 (see Figure 1) outputs volume information C4, which indicates the volume of solder printed on the substrate S1 through the opening H1 of the mask M1 in the previous (N-1th, where N is an integer of 2 or more) printing process using the mask M1. The acquisition unit 10 of the printing result prediction device 100 (see Figure 1) acquires the volume information C4.

[0027] The calculation unit 11 (see Figure 1) of the printing result prediction device 100 inputs printing conditions 16 into the prediction model 15 and predicts the volume of solder printed on the substrate S1 through the opening H1 of the mask M1 when the next (immediately following) printing process is performed using the mask M1 under the input printing conditions 16. That is, the calculation unit 11 inputs printing conditions 16 into the prediction model 15 and outputs a predicted value of the volume of solder printed on the substrate S1 through the opening H1 of the mask M1 when the Nth (next) printing process is performed using the mask M1. The prediction model 15 is a machine learning model that has been pre-trained using machine learning to calculate the volume of solder printed in the past.

[0028] As shown in Figure 4, the input printing conditions 16 include: (1) mask information C1 related to the mask M1 used in the printing process, (2) solder information C2 related to the solder used in the printing process, (3) device parameters C3 indicating the operation of the printer 110 in the printing process, (4) volume information C4 indicating the volume of solder printed through the opening H1 of the mask M1 in the previous printing process, and (5) printing sequence C5 indicating the number of times the mask M1 is used in the printing process. The printing conditions 16 may further include: (6) environmental information C6 of the printing process, and (7) substrate information C7 related to the substrate S1 used in the printing process.

[0029] The masking information C1 may include, for example, the thickness of the mask M1 and the opening size (opening area) of the opening H1. The masking information C1 may include the coordinates of the opening H1. The masking information C1 may include information about the processing method of the mask M1. The masking information C1 may include the angle (taper angle) of the side of the opening H1. The masking information C1 may include the material of the mask M1. The masking information C1 may include rigidity information such as the Young's modulus of the mask M1. The processing method may be, for example, laser processing or etching, a method of setting the opening H1 on the mask M1. For example, the arithmetic unit 11 reads the masking information C1 from the memory unit 13 (see Figure 1). Alternatively, the masking information C1 (or the model number of the mask M1 used, etc.) may be input to the arithmetic unit 11 from outside, such as the printing press 110, via the acquisition unit 10.

[0030] Solder information C2 may include, for example, information about the particle size of the solder. The particle size of the solder may be, for example, the average particle size of the solder used in the printing process. Solder information C2 may include information such as the particle size deviation of the solder, the viscosity of the solder, the thixotropic index, and the flux content. For example, the calculation unit 11 reads the solder information C2 from the memory unit 13. Alternatively, the solder information C2 (or the type of solder used, etc.) may be input to the calculation unit 11 from an external source such as the printing press 110 via the acquisition unit 10. Alternatively, the solder information C2 may be estimated separately using an estimation system.

[0031] Device parameter C3 may include, for example, the speed at which the squeegee 25 moves along the mask M1 (printing speed), the pressure at which the squeegee 25 contacts the mask M1 (printing pressure), the angle of the squeegee 25 relative to the mask M1 (squeegee angle), and the speed at which the substrate S1 is removed from the mask M1 (demolding speed). Device parameter C3 may include the distance between the mask M1 and the substrate S1 (gap). Device parameter C3 may include the frequency of automatic mask cleaning (cleaning frequency). Device parameter C3 may be, for example, a combination of multiple parameters such as printing speed and printing pressure.

[0032] The device parameter C3 input to the prediction model 15 is, for example, a candidate for the device parameter C3 in the next printing process. For example, the arithmetic unit 11 obtains the device parameter C3 from the printing press 110 or the memory unit 13. For example, the arithmetic unit 11 obtains a plurality of parameters representing the actions that the printing press 110 can perform from the printing press 110 or the memory unit 13, and selects the device parameter C3 (a candidate for the device parameter C3 in the next printing process) input to the prediction model 15 from the plurality of parameters obtained.

[0033] The volume information C4 is obtained by the acquisition unit 10 from the inspection machine 120. Furthermore, the range of "solder volume" input and output of the calculation unit 11 or the prediction model 15 includes values ​​of indicators corresponding to the solder volume. Specifically, for example, printing transfer rate or a value that can be converted into solder volume can be used. The printing transfer rate is the ratio of the volume of solder printed through the opening H (hole) of the mask to the volume of that opening H.

[0034] The printing sequence C5 represents the number of times the mask M1 is used in the printing process. That is, the input printing sequence C5 indicates which number of times the mask M1 will be used in the next printing process. In other words, the input printing sequence C5 indicates that the printing process using the mask M1 is the Nth time. For example, the calculation unit 11 obtains the number of times the mask M1 is used from the printing press 110 via the acquisition unit 10. Alternatively, a signal indicating the implementation of the printing process can be input from the printing press 110 to the calculation unit 11, and the calculation unit 11 can count the number of times the mask M1 is used.

[0035] Environmental information C6 may include, for example, the temperature or humidity inside the printing press 110. Environmental information C6 may also include the vibration of the printing press 110. The temperature or humidity inside the printing press 110 may be measured, for example, by a thermometer 111 or a hygrometer installed on the printing press 110. Alternatively, the temperature or humidity inside the printing press 110 may also be measured by a thermometer 111 or a hygrometer built into the printing press 110. The vibration of the printing press 110 may be measured, for example, by a vibration meter installed on the printing press 110. The processing unit 11 acquires the temperature or humidity measured by the thermometer 111 or hygrometer, and the vibration measured by the vibration meter, via the acquisition unit 10.

[0036] The substrate information C7 includes silkscreen information related to the silkscreen provided on the substrate S1. In this example, the silkscreen information is the image entropy calculated based on an image of the substrate S1 displaying the position or shape of the silkscreen. Details regarding the image entropy will be described later. The substrate information C7 is not limited to silkscreen information; it may also include the thickness or number of layers of the substrate S1, the thickness of the copper foil of pad P1, etc. The substrate information C7 may include the coating area of ​​the solder resist liquid coated on the surface of the substrate S1. The substrate information C7 may also include the warpage state of the substrate S1. For example, the substrate information C7, such as the silkscreen information, is input from an external device 112 (computer or memory device) that stores the design information of the substrate S1 to the processing unit 11 via the acquisition unit 10. Alternatively, for example, an image of the substrate S1 may be input from an external device 112 to the processing unit 11, and the image entropy is calculated based on the image acquired by the processing unit 11. The processing unit 11 may read the substrate information C7 or the image of the substrate S1 from the memory unit 13.

[0037] In addition, printing conditions 16 may include rigidity information such as the thickness, length, overhang, material, and Young's modulus of the squeegee 25 used in printing. Printing conditions 16 may also include the elapsed time after solder is replenished to the printing press 110, the amount of solder on the mask M1, and the tension of the mask M1. Printing conditions 16 may also include the type or position of the fixture supporting the substrate S1 for printing from below.

[0038] Figures 5(a) to 5(h) are schematic cross-sectional views illustrating the printing process. Figures 5(a) to 5(d) are examples of the N-1th printing process using the opening H1 of mask M1. As shown in Figure 5(a), the side surface W1 of the opening H1 of the mask M1 sometimes has irregularities. For example, depending on the processing method of the opening H1 when manufacturing the mask M1, the side surface W1 may become rough.

[0039] As shown in Figure 5(b), solder 20 is filled into the opening H1, and then, as shown in Figure 5(c), the mask M1 is removed from the substrate S1. Consequently, as shown in Figure 5(d), a portion 20a of the solder 20 filled into the opening H1 is printed onto the pad P1 of the substrate S1. Another portion 20b of the solder 20 filled into the opening H1 is attached to the side surface W1 of the mask M1. For example, the amount of the portion 20b of solder 20 attached to the side surface W1 varies depending on the roughness or unevenness of the side surface W1. Therefore, the volume of the solder 20 printed on the substrate S1 (the volume of a portion 20a) is considered to reflect the shape of the side surface W1. The volume information C4 described in Figure 4 is, for example, the volume of the portion 20a of solder 20 filled into the opening H1. Therefore, the volume information C4 reflects the roughness or unevenness of the side surface W1.

[0040] Figures 5(e) to 5(h) illustrate the Nth printing process using the opening H1 of the mask M1. In the Nth printing process, solder is printed on the pad P1 of a substrate S1 that is different from the N-1th printing process.

[0041] As shown in Figure 5(e), in the Nth printing process, a portion 20b of the solder 20 adhered in the N-1th printing process remains on the side W1 of the opening H1. For this opening H1, as shown in Figure 5(f), new solder 20 is filled, and then, as shown in Figure 5(g), the mask M1 is removed from the substrate S1. Thereby, as shown in Figure 5(h), a portion 20c of the solder 20 filled in the opening H1 is printed onto the substrate S1. An amount of solder 20 corresponding to the roughness or unevenness of the side W1 is adhered to the side W1 of the opening H1.

[0042] In this way, based on the volume of solder 20 printed in the N-1th printing process (i.e., volume information C4), information on the roughness, unevenness, and other shapes of the side surface W1 of the opening H1 can be obtained. Furthermore, the volume of solder 20 (a portion 20c of the volume) printed on the pad P1 of the substrate S1 in the Nth printing process varies depending on the shape of the side surface W1.

[0043] Figures 6(a) to 6(h) are schematic cross-sectional views illustrating the printing process. Figures 6(a) to 6(d) illustrate the (N-1)th printing process using the opening H1 of the mask M1. As shown in Figure 6(a), the mask M1 is placed on the substrate S1. As shown in Figure 6(b), solder 20 is filled into the opening H1 of the mask M1. As shown in Figure 6(c), the mask M1 is removed from the substrate S1. Consequently, as shown in Figure 6(d), a portion 20e of the solder 20 filled into the opening H1 is printed onto the pad P1 of the substrate S1. The remaining portion 20f of the solder 20 filled into the opening H1 is attached to the side W1 of the mask M1.

[0044] Figures 6(e) to 6(h) illustrate an example of the Nth printing process using the opening H1 of the mask M1. As shown in Figure 6(e), the mask M1 is disposed on the substrate S1. In the Nth printing process, a portion 20f of the solder 20 attached in the (N-1)th printing process remains on the side W1 of the opening H1. For such an opening H1, as shown in Figure 6(f), new solder 20 is filled, and then, as shown in Figure 6(g), the mask M1 is removed from the substrate S1. Thereby, as shown in Figure 6(h), a portion 20g of the solder 20 filled in the opening H1 is printed onto the substrate S1. Solder 20 is further attached to the side W1 of the opening H1.

[0045] In this way, by repeating the printing process, the volume of solder 20 attached to the side W1 of the opening H1 changes. For example, if the number of printing processes, that is, the number of times the mask M1 is used, increases, the volume of solder 20 attached to the side W1 increases, and the volume of solder 20 printed on the substrate S1 decreases.

[0046] Based on the number of times the mask M1 is used (i.e., the printing sequence C5), information such as the state of the side W1 and the amount of solder 20 attached to the side W1 can be obtained. Moreover, the volume of solder 20 printed on the substrate S1 through the opening H1 changes due to the state of the side W1.

[0047] As described above, in the embodiment, the prediction model 15, in addition to mask information C1, solder information C2, and device parameters C3, also inputs volume information C4 and printing sequence C5. This allows for the prediction of the printed solder volume by considering, for example, the shape (unevenness, roughness) of the side surface W1 of the mask M1, and the condition of the side surface W1 (e.g., the amount of solder adhering to it). According to the embodiment, the prediction accuracy of the printed solder volume can be improved.

[0048] For example, it is difficult to directly measure the shape of the side W1 of the mask M1 or the amount of solder attached to the side W1. In contrast, in the embodiment, by using volume information C4 and printing sequence C5, information related to the state of the side W1 of the mask M1 can be easily obtained.

[0049] Furthermore, for example, by cleaning the mask M1, the solder adhering to the side W1 is removed. The number of times the mask M1 is used as shown in the printing sequence C5, for example, the number of times the mask M1 is used after cleaning, is reset to zero by cleaning the mask M1.

[0050] Figure 7 is a schematic diagram illustrating the image entropy containing silkscreen information. For example, the silkscreen information contained in the substrate information C7 illustrated in Figure 4 is based on the image IM1 of the substrate S1. Image IM1 displays the image of pad P1 and the surrounding area of ​​pad P1.

[0051] In this example, the screen printing information refers to the information of the screen printing Sk around the pad P1. That is, the screen printing information is related to the position and shape of the screen printing around the pad on which solder is printed through the opening of the mask used in the printing process.

[0052] Furthermore, the perimeter of the solder pad refers to the specified size range of the perimeter of the pad. For example, the perimeter of the solder pad is a square area approximately 1-10 cm in size with one side centered on the center of the pad. Image IM1 can be, for example, an image obtained by photographing substrate S1, or a design drawing of substrate S1.

[0053] For example, the silkscreen information is the image entropy of the image showing the silkscreen Sk information surrounding the display pad P1. For example, as shown in Figure 7, an image IM2 that captures only the silkscreen Sk image is derived from image IM1. Binarization can be applied to image IM2. Binarization sets the pixels displaying the silkscreen Sk image to black and all other pixels to white. Furthermore, for example, an image IM3 obtained by applying multiple dilation processes to image IM2 can be derived. The dilation process replaces the pixels adjacent to the black-pixelated silkscreen Sk image in image IM2 with black pixels.

[0054] As screen printing information, the image entropy of image IM3 can be used. Image entropy (H) is calculated according to the formula H = -ΣPilog2Pi. i represents the subscript of the pixel value order. Pi represents the probability of occurrence of the i-th order pixel in the image. Based on the sum of i, the image entropy (H) is calculated.

[0055] Furthermore, the dilation process for the exported image IM3 can be omitted as needed. For example, the image entropy of image IM2 can be used as screen printing information.

[0056] Figure 8 is a schematic cross-sectional view illustrating the printing process performed by a printing press. As shown in Figure 8, in the printing process, for example, a mask M1 is disposed on a substrate S1 with screen printing Sk. The substrate S1 is supported by a plurality of pins 115 at a plurality of positions.

[0057] The silkscreen Sk is a protrusion extending upward from the surface of the substrate S1. The silkscreen Sk is located between the mask M1 and the substrate S1. When the silkscreen Sk is present, the substrate S1 may sometimes bend due to the force from a fixture such as pin 115. In the case of substrate S1 bending, the distance between each pad P and the substrate S1 varies for each pad P depending on the position or shape of the silkscreen Sk around each pad P. As a result, the amount of solder 20 filling each opening H on each pad P varies. That is, depending on the position or shape of the silkscreen Sk, the volume of solder printed on the pad P may sometimes vary.

[0058] In contrast, in the implementation, the printing conditions 16 input to the prediction model 15 can include screen printing information as described above. By taking into account the screen printing information, the volume of solder to be printed can be predicted, thereby further improving the prediction accuracy.

[0059] Figure 9 is a bar chart illustrating the accuracy of the predicted volume of printed solder. When calculating the proportion of the achieved data shown in Figure 9, the prediction model 15 is used to predict the volume of each solder piece printed on each pad P through each opening H of the mask M. Furthermore, the volume of each solder piece printed on each pad P through each opening H is measured. The proportion of the achieved data is the ratio of the number of values ​​with prediction errors smaller than the specified value (specification value) to the total number of predicted values. The prediction error corresponds to the difference between the measured value of the solder volume and the predicted value of the solder volume.

[0060] Figure 9 shows the data for the case where printing condition 16 in prediction model 15 does not include silkscreen information. "Distance," "Amount," and "Entropy" are the data for the case where printing condition 16 in prediction model 15 includes silkscreen information. In "Distance," the silkscreen information is the distance between pad P and the silkscreen Sk closest to pad P. In "Amount," the silkscreen information is the amount (e.g., area) of the silkscreen Sk surrounding pad P. In "Entropy," the silkscreen information is the image entropy as illustrated in Figure 7.

[0061] As shown in Figure 9, prediction accuracy can be further improved by considering silkscreen information. Prediction accuracy can be further improved when the silkscreen information includes image entropy calculated based on the image of the substrate.

[0062] Figure 10 is a schematic diagram illustrating a variation in the prediction made by the printing result prediction device in the implementation mode. In the example of Figure 4 above, image entropy is used in the silkscreen information of substrate information C7. The example of Figure 10 differs from the example described in Figure 4 in that the image of substrate S1 is used instead of image entropy in substrate information C7.

[0063] As shown in Figure 10, as substrate information C7, an image IM1 can also be input, showing the pad P1 for solder printing and its surrounding area, which is the predicted object for volume. Alternatively, an image IM2 showing only the pad P1 and the silkscreen Sk can be generated based on the image IM1, and this image IM2 can be input as substrate information C7. The image IM2 can be generated, for example, by a machine learning model 19 that learns by taking the image IM1 as input and outputting the image IM2.

[0064] Figure 11 is a schematic diagram illustrating the prediction made by the result prediction device in the implementation mode. The example shown in Figure 11 differs from the example illustrated in Figure 4 in that it uses the bump information C70 as the substrate information C7.

[0065] A plurality of protrusions Cp are provided on the surface of substrate S1 (see Figure 12, etc.). The bump information C70 is information related to the protrusions Cp provided on substrate S1. More specifically, the bump information includes at least one of the following: thickness information C71 related to the thickness of the protrusion Cp; distance information C72 corresponding to the distance between the protrusion Cp and pad P1; and image entropy C73 calculated based on an image containing the protrusions Cp. The protrusions Cp may, for example, include a silkscreen print Sk. The following explanation will use the case where the protrusion Cp is a silkscreen print Sk as an example.

[0066] Figure 12 is a schematic plan view illustrating a portion of the substrate. As shown in Figure 12, a plurality of protrusions Cp (screened Sk in this example) are arranged around the pad P1. Distance information C72 is, for example, the shortest distance Dm between the center P1c of the pad P1 and the protrusions Cp. In other words, distance information C72 is the distance between the center P1c and the protrusion Cp closest to the center P1c. Distance information C72 is not limited to the shortest distance Dm itself; sometimes it can also be a distance corresponding to the shortest distance Dm, and for example, the distance between the pad P1 and the protrusion Cp closest to the pad P1. For example, distance information C72 can sometimes also be the shortest distance between the end of the pad P1 and the protrusion Cp.

[0067] Distance information C72 can be obtained, for example, from the design information of substrate S1. For example, the shortest distance Dm can be calculated based on the position or shape of silkscreen Sk (e.g., a binarized image of silkscreen Sk on the substrate) and the coordinates of pad P1, according to the design information of substrate S1.

[0068] Figure 13(a) and Figure 13(b) are schematic plan views illustrating a portion of the substrate, and Figure 13(c) is a curve diagram illustrating the unevenness of the substrate. A plurality of pads, silkscreened markings (Sk), or wirings are disposed on the surface of substrate S1, which are omitted for simplicity in Figure 13(a). Figure 13(b) is an enlarged view of one of the plurality of regions R1 shown in Figure 13(a). On the substrate surface, along the straight line L1 shown in Figure 13(b), unevenness is generated due to the protrusion Cp. Figure 13(c) is an example of unevenness along the straight line on the substrate surface measured by a three-dimensional measuring machine.

[0069] For example, as shown in Figure 13(c), the thickness information C71 is the thickness TCp of the protrusion Cp (the height of the silkscreen Sk). The thickness TCp of the protrusion Cp can be based on measured values ​​or on the design information of the substrate S1.

[0070] The thickness information C71 is not limited to the thickness of the protrusion Cp closest to pad P1 itself, but can also be a value corresponding to the thickness of the protrusions Cp disposed around pad P1. For example, each of the plurality of regions R1 on the substrate S1 contains one or more protrusions Cp. The thickness information C71 can be the average of the thicknesses (measured values) of the plurality of protrusions Cp contained in the plurality of regions R1. In this example, the thickness information C71 is the average thickness of the plurality of protrusions Cp in the three regions R1. Furthermore, there are no particular limitations on the size of a region R1, the position of region R1, or the number of regions R1, as long as they are appropriately determined.

[0071] Image entropy C73 can be calculated from the image obtained by capturing the protrusion Cp from the image of the substrate S1 using the same method as described in FIG. 7 above. For example, in the case where the protrusion Cp is a silkscreen Sk, image entropy C73 is the image entropy described with reference to FIG. 7.

[0072] Figure 14 is a schematic diagram illustrating the substrate and mask in the printing process. As shown in Figure 14, when a mask M1 is disposed on a substrate S1, the contact between the substrate S1 and the mask M1 changes in response to the unevenness caused by the protrusion Cp on the substrate S1. For example, the distance between the pad P1 and the mask M1 changes. Therefore, based on the unevenness information C70 related to the protrusion Cp disposed on the substrate S1, the volume of the printed solder changes.

[0073] Figure 15 is a graph illustrating the relationship between the distance between the protrusion and the pad and the rate of increase in solder volume. In Figure 15, the increase rate is the percentage increase in the volume of solder printed on pad P1, based on the case without the raised portion Cp (or the case where the raised portion Cp is sufficiently far away from pad P1). Figure 15 shows data for the case where the thickness of the raised portion Cp (silk screen Sk) is 21 μm and data for the case where the thickness of the raised portion Cp (silk screen Sk) is 81 μm.

[0074] As shown in Figure 15, the volume of printed solder increases when the distance between the protrusion Cp and the pad P1 becomes shorter. In particular, the rate of increase in volume is even greater when the protrusion Cp is thicker.

[0075] In contrast, in the example of Figure 11, printing condition 16 includes substrate information C7 containing bump and recess information C70. By taking into account the bump and recess information C70, the volume of the printed solder can be predicted, thereby improving the prediction accuracy.

[0076] Figure 16 is a schematic diagram illustrating another example of the prediction performed by the result prediction device in the implementation mode. In this example, the printing condition 16, which serves as the input to the prediction model 15, does not include volume information C4 and printing order C5, which differs from the example illustrated in Figure 11.

[0077] In this example, printing condition 16 also includes substrate information C7 containing bump and recess information C70. This improves prediction accuracy. Furthermore, even if printing condition 16 does not include volume information C4, since the previous printing results are not required during printing, the solder volume can be predicted with higher accuracy, even in the first printing.

[0078] Figure 17 is a bar chart illustrating the prediction error of the volume of printed solder. "(A) Distance + Thickness", "(B) Distance", "(C) Thickness", and "(D) None" respectively represent the cases in which the input of prediction model 15 includes thickness information C71 and distance information C72, includes distance information C72, includes thickness information C71, and does not include bump information C70. Furthermore, the input of each prediction model 15 includes masking information C1, solder information C2, and device parameters C3, but does not include volume information C4 or printing order C5.

[0079] In the case of "(A) distance + thickness", the prediction error is about 40% smaller than that in the case of "(D) none". When the prediction model 15 is a model that takes printing condition 16 containing the embossing information C70 as input, the prediction error can be reduced compared to the case of a model that does not include the embossing information C70 in the input.

[0080] Figure 18 is a schematic plan view illustrating a portion of the substrate. An outer layer circuit can be disposed in the substrate S1. As shown in Figure 18, the outer layer circuit includes wiring Wa that protrudes upward from the surface of the substrate. In the above description, the case where the raised portion Cp is the silkscreen Sk was used as an example, but the raised portion Cp is not necessarily limited to the silkscreen Sk. For example, the raised portion Cp can also be the wiring Wa provided on the surface of the substrate S1. Alternatively, the intersection of the silkscreen Sk and the wiring Wa (the part where the silkscreen Sk and the wiring Wa overlap in the thickness direction perpendicular to the substrate surface) can be defined as the raised portion Cp.

[0081] For example, both the silkscreen Sk and the wiring Wa can be designated as protrusions Cp. That is, a plurality of protrusions Cp can include a plurality of first protrusions (e.g., silkscreen Sk) and a plurality of second protrusions (e.g., wiring Wa). In this case, the thickness information C71 can include at least one of first thickness information related to the thickness of the first protrusion and second thickness information related to the thickness of the second protrusion. The distance information C72 can include at least one of first distance information corresponding to the distance between the pad and the first protrusion and second distance information corresponding to the distance between the pad and the second protrusion. The image entropy can be calculated based on images of at least one of the first protrusion and the second protrusion. Alternatively, the convexity / concaveness information 70 can include a first image entropy calculated based on the image of the first protrusion and a second image entropy calculated based on the image of the second protrusion.

[0082] Figure 19 is a schematic diagram illustrating the learning process of the prediction model. The printing result prediction device 100 may include a learning unit 17 that performs machine learning on the prediction model 15. The learning unit 17 generates the prediction model 15 in advance by means of machine learning before predicting the solder volume.

[0083] The learning unit 17 is, for example, a computer equipped with a deep learning program. The prediction model 15 is, for example, a model learned using a neural network or random forest. The prediction model 15 learns by taking printing conditions as input and outputting a predicted value of the volume of the printed solder. In machine learning, well-known techniques can be appropriately applied.

[0084] Specifically, for example, one or more masks M2 are used. Multiple printing processes are performed in advance. That is, the printing machine 110 performs a printing process in which solder is printed on the pad P2 of the substrate S2 by filling the opening H2 of the mask M2 disposed on the substrate S2 with solder. Furthermore, the substrate S2, mask M2, opening H2, and pad P2 are examples of the substrate S, mask M, opening H, and pad P described in Figures 2(a) to 3(c). The mask M2 and substrate S2 used in the advance printing process can be the same as (or of the same specifications) as the mask M1 and substrate S1 used in the printing process of the object of the printing result prediction (the printing process of solder with predicted printing volume) described in Figure 4, or they can be different.

[0085] Learning Unit 17 studies the printing conditions in each prior printing process and the relationship between the volume of solder printed through each opening H2 of each mask M2 used in each prior printing process. That is, for example, Learning Unit 17 uses the printing conditions in the nth printing process (n is an integer greater than 2) performed through the opening H2 of mask M2 and the volume Vn of solder printed through the opening H2 of mask M2 in the nth printing process using mask M2 as teaching data for learning.

[0086] Furthermore, the printing conditions related to the nth printing process using mask M2 may include, for example: (1) mask information C1 related to mask M2, (2) solder information C2 related to the solder used in the nth printing process using mask M2, (3) device parameters C3 indicating the operation of the printer 110 in the nth printing process using mask M2, (4) volume information C4 indicating the volume of solder printed through the opening H2 of mask M2 in the (n-1)th printing process using mask M2, and (5) printing sequence C5 indicating the number of times mask M2 is used in the nth printing process using mask M2 (in other words, information indicating that the printing process using mask M2 is the nth). Furthermore, the printing conditions related to the nth printing process using mask M2 may include: (6) environmental information C6 of the printing process, and (7) substrate information C7 related to the substrate S2 used in the printing process. For example, substrate information C7 may include bump and recess information C70.

[0087] As illustrated in Figure 16, printing condition 16 may omit volume information C4 and printing sequence C5, but include substrate information C7 containing bump and relief information C70. In this case, for example, the learning unit 17 learns the relationship between each printing condition in a prior plurality of printing processes and the volume of solder printed by each opening H2 of each mask M2 in the printing process under each printing condition. The printing conditions in the prior printing processes include, for example: (1) mask information C1 related to the mask M2 used in the printing process, (2) solder information C2 related to the solder used in the printing process, (3) device parameters C3 indicating the operation of the printer 110 in the printing process, and (4) substrate information C7 containing bump and relief information C70 of the substrate S2 used in the printing process. In the case of performing a plurality of printing processes under one printing condition, the learned solder volume may be the average value of the plurality of printing processes.

[0088] Figure 20 is a schematic diagram illustrating the operation of the printing system in its implementation form. As shown in Figure 20, the acquisition unit 10 of the printing result prediction device 100 (refer to Figure 1) acquires inspection information Z from the inspection machine 120. The inspection information Z includes volume information C4 (refer to Figure 4) of the solder volume printed in the previous printing process. Furthermore, the acquisition unit 10 of the printing result prediction device 100 acquires environmental information C6, such as temperature. Then, information w is input into the printing result prediction device 100. Information w includes masking information C1, solder information C2, and substrate information C7, as explained in Figure 4.

[0089] Furthermore, the calculation unit 11 of the printing result prediction device 100 acquires the device parameters C3 and printing sequence C5 as described in FIG4. Based on the information w, device parameters C3, volume information C4 and printing sequence C5, the calculation unit 11 predicts the volume of solder to be printed in the next printing process using a pre-trained prediction model 15.

[0090] The calculation unit 11 inputs a plurality of printing conditions 16 into the prediction model 15 to predict the volume of solder to be printed when the next printing process is performed under each of the input printing conditions 16. The plurality of printing conditions 16 input into the prediction model 15 are conditions that cause the device parameter C3 to change. That is, for example, the plurality of printing conditions 16 used in predicting the volume of solder printed through the same opening H are different in the device parameter C3, while the other conditions are the same.

[0091] For example, the calculation unit 11 calculates the print quality y with the device parameter C3 set as variable x based on the prediction result. The calculation unit 11 determines the device parameter C3 for the next printing process based on the change in print quality y when the device parameter C3 (variable x) is changed, and outputs it to the printing press 110.

[0092] In this way, based on the predicted solder volume, the device parameter C3 for the next printing process is determined. The printer 110 performs the next printing process by means of the action indicated by the device parameter C3 for the next printing process determined by the computing unit 11.

[0093] More specifically, the calculation unit 11 predicts the volume of solder printed onto a plurality of pads P using each of the plurality of openings H provided on the mask M1. For example, the acquisition unit 10 acquires a plurality of volume information C4 representing the volume of solder printed using each of the plurality of openings H of the mask M1 in the previous printing process using the mask M1. The calculation unit 11 inputs printing conditions 16 to the prediction model 15 for each of the plurality of openings H, and predicts the volume of solder to be printed when performing the next printing process under each of the input plurality of printing conditions 16. Furthermore, among the plurality of printing conditions 16 (a plurality of device parameters C3) input to the prediction model 15, a plurality of device parameters C3 common to the plurality of openings H are used. Then, based on the prediction result, the calculation unit 11 determines, for example, the device parameter C3 contained in one of the plurality of printing conditions 16 as the device parameter C3 for the next printing process.

[0094] Print quality y is, for example, a parameter representing the overall print quality of solder printed on a plurality of pads P present on a substrate S1. More specifically, print quality y can include, for example, the "number of pads printed with solder that is judged to be acceptable". Whether a product is acceptable is determined by whether the volume of the printed solder meets the specified criteria. The specified criteria are, for example, a print transfer rate of 95% or more and 100% or less, but are not limited to this and can be appropriately determined. Alternatively, print quality y can include, "the sum of the deviations of the print transfer rates of each pad on the substrate from 100%". That is, the absolute value of the difference between the print transfer rate (%) of the pad and 100% can be used as the deviation of the print transfer rate of the pad, and the sum of the deviations of the print transfer rates of a plurality of pads can be used as print quality y.

[0095] In methods for determining the device parameter C3 to be used in the next printing process based on print quality y, methods such as global search or gradient descent can be used. In the global search method, the print quality y for all device parameters C3 that can be set by the printing press 110 is calculated, and the device parameter C3 that best represents print quality y (e.g., the device parameter C3 that maximizes print quality y) is selected. In the gradient descent method, the print quality y is first calculated using a provisional value of device parameter C3 as input. Based on the gradient of print quality y as device parameter C3 changes from the provisional value, the provisional value of device parameter C3 is updated to improve print quality y. By repeating this update, the device parameter C3 that achieves a print quality y better than a specified reference value is calculated.

[0096] Furthermore, the aforementioned method of determining the equipment parameter C3 to be used in the next printing process based on the print quality y is just one example and is not necessarily limited to the above.

[0097] According to the printing system 200 in the implementation form, for example, in the production of a substrate printed with solder, the device parameter C3 can be appropriately updated in response to changes in temperature or changes in the state of the side of the opening H of the mask M. For example, the appropriate device parameter C3 can be automatically calculated based on information about the object to be printed. The printing result prediction device 100 predicts the printing quality, for example, even without setting conditions to operate the actual production line, the appropriate device parameter C3 can be set. Furthermore, for example, the appropriate device parameter C3 can be set without relying on the user's experience. For example, the time required to set the device parameter C3 can be shortened.

[0098] Figure 21 is a schematic diagram illustrating the configuration of a printing result prediction device in an exemplary embodiment. The printing result prediction device 100 described above, as an example, is implemented using hardware identical to that of a general computer (information processing device). The printing result prediction device 100 includes: a CPU (Central Processing Unit) 311, an input device 312, a display device 313, a ROM (Read Only Memory) 314, a RAM (Random Access Memory) 315, a memory device 316, a communication device 317, and a bus 318. All components are connected via the bus 318.

[0099] The CPU 311 collaborates with various programs pre-memorized in the ROM 314 or memory device 316 to perform various processes and coordinate the operation of each component constituting the printing result prediction device 100. During processing, the CPU 311 uses a designated area of ​​the RAM 315 as its operating area. The CPU 311 collaborates with the programs pre-memorized in the ROM 314 or memory device 316 to implement the input device 312, display device 313, and communication device 317, etc.

[0100] Input device 312 includes, for example, at least one of a keyboard, a mouse, and a touch panel. Input device 312 accepts information input by the user as an indication signal and outputs the indication signal to CPU 311. Display device 313 is, for example, a monitor. Display device 313 outputs various information in a visually readable manner based on the signals output from CPU 311.

[0101] ROM 314 stores the programs and various settings used in the control of the printing result prediction device 100 in a non-rewritable manner. RAM 315 is a volatile memory medium such as SDRAM (Synchronous Dynamic Random Access Memory). RAM 315 functions as the operating area of ​​CPU 311. Specifically, it functions as a buffer for temporarily storing various variables and parameters used by the printing result prediction device 100.

[0102] Memory device 316 is a rewritable recording device based on semiconductor memory media such as flash memory, or memory media that can be magnetically or optically recorded. Memory device 316 stores the program and various setting information used in the control of printing result prediction device 100, pre-trained prediction models, and a database containing model names and specifications with masks, etc. Communication device 317 is used to communicate with external machines to send and receive information.

[0103] For example, CPU 311, ROM 314 and RAM 315 function as arithmetic unit 11 (see Figure 1), input device 312 and communication device 317 function as acquisition unit 10 (see Figure 1), and memory device 316 function as memory unit 13 (see Figure 1).

[0104] The implementation form may include the following components (e.g., technical solutions). (Component 1) A printing result prediction device for predicting the volume of solder to be printed by a printing press, the printing press printing solder on the substrate by filling an opening in a mask disposed on the substrate, and the printing result prediction device comprising: The computing unit inputs printing conditions into a prediction model pre-trained by machine learning, and predicts the volume of solder to be printed using the aforementioned masking and opening when performing the next printing under the input printing conditions; and The aforementioned printing conditions include: Information related to the use of the aforementioned masks in printing. Solder information related to solder used in printing. Substrate information, which includes bump information related to the protrusions of the aforementioned substrate used in printing, and The device parameters represent the operation of the aforementioned printing machine in the aforementioned printing process. (Component 2) For example, in the printing result prediction device constituting 1, the aforementioned prediction model is a model learned by neural networks or random forests. (Component 3) As in the printing result prediction device constituting 1 or 2, the aforementioned substrate used in printing has a pad for printing solder through the aforementioned opening. The aforementioned convexity / concave information includes at least one of the following: thickness information related to the thickness of the aforementioned convex portion, distance information corresponding to the distance between the aforementioned convex portion and the aforementioned pad, and image entropy calculated based on an image containing the aforementioned convex portion. (Component 4) If the printing result prediction device constitutes any one of 1 to 3, wherein the aforementioned protrusion includes screen printing. (Component 5) If a printing result prediction device is configured as any one of 1 to 4, wherein the aforementioned protrusion includes wiring of the outer layer circuit. (Composition 6) If a printing result prediction device is configured as any one of 3 to 5, wherein the aforementioned thickness information is the average thickness of the aforementioned protrusions in a plurality of regions disposed on the aforementioned substrate. (Component 7) If a printing result prediction device is configured as any one of 3 to 6, the aforementioned distance information is the shortest distance between the center of the aforementioned pad and the aforementioned protrusion. (Composition 8) If the printing result prediction device constitutes any one of 1 to 7, it further includes an acquisition unit that acquires volume information indicating the volume of solder printed through the aforementioned opening of the aforementioned mask during the previous printing; and The aforementioned printing conditions include: The aforementioned volume information indicates the volume of solder printed during the previous printing, and This indicates the printing sequence in which the aforementioned masking is used during printing. (Composition 9) A printing system comprising: a printing result prediction device as configured in any of 1 to 8, The aforementioned printing press, and The inspection machine checks the aforementioned volume of the printed solder; and The aforementioned acquisition unit obtains from the aforementioned inspection machine the aforementioned volume information of the solder printed during the previous printing. Based on the predicted solder volume, the aforementioned calculation unit determines the aforementioned device parameters for the next printing. The aforementioned printing press performs the aforementioned next printing based on the actions determined by the aforementioned calculation unit and indicated by the aforementioned device parameters for the next printing.

[0105] Depending on the implementation, a printing result prediction device and printing system can be provided that can improve the prediction accuracy of the volume of solder during printing.

[0106] The above examples illustrate several embodiments of the present invention, but these embodiments are provided as examples and are not intended to limit the scope of the invention. These novel embodiments can be implemented in various other forms, and various omissions, substitutions, and modifications can be made without departing from the spirit of the invention. These embodiments or variations thereof are included within the scope and spirit of the invention, and are also included within the scope of the invention described in the patent application and its equivalents. Furthermore, the aforementioned embodiments can be combined with each other.

[0107] 10: Obtaining Department 11: Arithmetic Department 13: Memory Department 15: Predictive Model 16: Printing conditions 17: Study Department 19: Machine Learning Models 20: Solder 20a~20g: Part of the solder 25: Scraper 100: Printing Result Prediction Device 110: Printing press 111: Thermometer 112: Device 115: Sales 120: Inspection machine 200: Printing System 311: CPU 312: Input device 313: Display device 314:ROM 315: RAM 316: Memory device 317: Communication devices 318: Busbar B: Substrate C1: Masking Information C2: Solder Information C3: Device Parameters C4: Volume Information C5: Printing order C6: Environmental Information C7: Substrate Information C70: AoTu Information C71: Thickness Information C72: Distance Information C73: Image Entropy Cp: ​​convex part Dm: Shortest distance H, H1, H2: Openings IM1, IM2, IM3: Images L1: Line M, M1, M2: Mask P, P1, P2: Pads P1c: Center R1: Region S, S1, S2: Substrate Sk: Silkscreen TCp: Thickness Vn: Volume w: News W1: Side view Wa: Wiring x: variable y: Print quality Z: Check Information

Claims

1. A printing result prediction device for predicting the volume of solder to be printed by a printing press, the printing press printing solder on the substrate by filling an opening of a mask disposed on the substrate with solder, and the printing result prediction device comprising: a computing unit that inputs printing conditions to a prediction model pre-trained by machine learning, and predicts the volume of solder to be printed through the opening of the mask when the next printing is performed under the input printing conditions; and the printing conditions comprising: mask information related to the mask used in printing, solder information related to the solder used in printing, and substrate information including protrusion information related to a protrusion disposed on the substrate used in printing, and device parameters indicating the operation of the printing press during printing.

2. The printing result prediction device as claimed in claim 1, wherein the aforementioned prediction model is a model learned by a neural network or random forest.

3. The printing result prediction apparatus of claim 1, wherein the substrate used in printing has a pad for printing solder through the aforementioned opening, and the aforementioned bump information includes at least one of thickness information related to the thickness of the aforementioned bump, distance information corresponding to the distance between the aforementioned bump and the aforementioned pad, and image entropy calculated based on an image including the aforementioned bump.

4. The printing result prediction device as claimed in any of claims 1 to 3, wherein the aforementioned protrusion includes screen printing.

5. The printing result prediction device as claimed in any of claims 1 to 3, wherein the aforementioned protrusion includes wiring of the outer layer circuit.

6. The printing result prediction device as claimed in claim 3, wherein the aforementioned thickness information is the average thickness of the aforementioned protrusions in a plurality of regions disposed on the aforementioned substrate.

7. The printing result prediction device as claimed in claim 3, wherein the aforementioned distance information is the shortest distance between the center of the aforementioned pad and the aforementioned protrusion.

8. The printing result prediction device according to any one of claims 1 to 3, further comprising an acquisition unit that acquires volume information indicating the volume of solder printed through the aforementioned opening of the aforementioned mask during the previous printing; and the aforementioned printing conditions include: the aforementioned volume information indicating the aforementioned volume of solder printed during the aforementioned previous printing, and the printing sequence indicating the number of times the aforementioned mask is used during printing.

9. A printing system comprising: a printing result prediction device as described in claim 8, the aforementioned printing press, and an inspection machine, wherein the aforementioned acquisition unit acquires from the aforementioned inspection machine the aforementioned volume information of the aforementioned volume of solder printed during the aforementioned previous printing, the aforementioned calculation unit determines the aforementioned device parameters for the aforementioned next printing based on the predicted aforementioned volume of solder, and the aforementioned printing press performs the aforementioned next printing according to the action determined by the aforementioned calculation unit and indicated by the aforementioned device parameters for the aforementioned next printing.

Images