Ai-based real-time process control system for multi-type dispensing

Abstract

The present invention relates to an AI-based real-time process control system for multi-type dispensing, comprising: a module identification unit that identifies the type of dispensing module that is exchangeably mounted on a dispensing equipment and outputs module identification information; an image vision unit that captures a coating result of a fluid coated on a substrate and generates a coating result image; an AI processing unit that includes a multi-type AI inspection model that has learned a set of process parameters, a coating result image, a good / bad status, and a defect type for each of the multi-type dispensing modules, and switches to an AI inspection model corresponding to the module identification information output from the module identification unit, and outputs a corrected set of process parameters according to a pre-learned defect type when the coating quality is determined to be defective based on the coating result image obtained from the image vision unit; and a system control unit that repeats the process of adjusting the process parameters of the dispensing equipment in real time according to the corrected set of process parameters of the AI ​​processing unit during the process.

Description

Technology Field

[0001] The present invention relates to precision fluid dispensing equipment, and more specifically, to an AI-based real-time process control system for multi-type dispensing that inspects the quality of a fluid applied on a substrate in real-time based on AI in response to multi-type dispensing methods such as jetting, progressive cavity pump, auger, vacuum, and time-pressure, and adaptively corrects process parameters in real-time by fusing the inspection results with viscosity change and residual amount data. Background Technology

[0003] In the field of precision fluid dispensing, various fluids such as solder paste, adhesives, underfills, silicones, and thermal conductive gels are applied onto substrates in the form of dots, beads, or fills. Since the quality of the application directly affects the electrical connection reliability, mechanical strength, and thermal performance of the product, determining the quality of the application results is very important.

[0004] In particular, in high-precision processes such as semiconductor packaging, automotive component assembly, and display module manufacturing, geometric characteristics of the coating results, such as bead width, height, volume, continuity, and pattern alignment, act as key factors determining product yield. Accordingly, there is a rapidly increasing demand for real-time quality inspection and automatic correction technologies for dispensing processes.

[0005] As an example of a real-time quality inspection and automatic correction technology for such a dispensing process, U.S. Patent No. 12233429 (February 25, 2025), cited as a prior art in this document, discloses a technology in which, in the prior art, quality inspection is performed offline through separate inspection equipment (AOI, SPI, etc.) after application, or, specialized for the jetting method, an image is taken of the fluid while it is in flight from the nozzle to the substrate to classify defects and adjust parameters.

[0006] However, conventional offline inspection methods have a structure that detects defects retrospectively after they occur, so problems are recognized only after defective products have been transferred to the next process, resulting in significant process losses.

[0007] In addition, there is a problem in that real-time parameter feedback is impossible because the offline inspection equipment is physically separated from the coating equipment.

[0008] Furthermore, conventional technologies utilizing in-flight images have a problem in that they are not suitable for inspecting the shape of the resulting product actually applied to the substrate, as they focus on capturing images of the fluid as it flies from the nozzle to the substrate. For example, since PCP and auger methods apply fluid in the form of continuous beads, it is technically appropriate to inspect the shape of the applied product landed on the substrate rather than the in-flight image; however, relying on images of the in-flight state leads to inaccuracies.

[0009] In addition, conventional technology utilizing images during flight has the problem that it is limited to defect types specific to the jetting method, such as satellite defects, nozzle accumulation defects, and misplacement defects, and therefore cannot be applied to other methods.

[0010] Fluid viscosity continuously fluctuates depending on temperature, usage time, and residual volume. Although these viscosity changes directly affect coating quality, there is a problem in that it is difficult to accurately analyze the causes of defects resulting from viscosity changes and to make preventive corrections because only vision inspection results are used as feedback input. Prior art literature

[0012] Patent Document 1: Registered Patent No. 10-2466306 (Registration Date: Nov. 08, 2022) Patent Document 2: Registered Patent No. 10-2330271 (Registration Date: Nov. 18, 2021) Patent Document 3: Registered Patent No. 10-2681795 (Registration Date: July 01, 2024) Patent Document 4: Registered Patent No. 10-2798555 (Registration Date: April 16, 2025) Patent Document 5: U.S. Patent No. 12233429 (Registration Date: Feb. 25, 2025) The problem to be solved

[0013] The present invention aims to provide an AI-based real-time process control system for multi-type dispensing that can determine the quality of a product coated on a substrate in real time, regardless of the coating method, in order to solve the aforementioned problems.

[0014] In addition, the purpose is to provide an AI-based real-time process control system for multi-type dispensing that has an adaptive inspection function that automatically recognizes the type of installed dispensing module and automatically loads an AI inspection model optimized for that method.

[0015] In addition, the aim is to provide an AI-based real-time process control system for multi-type dispensing that has a multi-sensor fusion closed-loop function that performs automatic parameter correction based on complex cause analysis by fusing vision inspection data, viscosity correction data, and residual amount data.

[0016] In addition, the purpose is to provide an AI-based real-time process control system for multi-type dispensing that is easy to retrofit to existing dispensing equipment by providing an independent vision inspection module that is detachable from the standard mounting interface of the dispensing equipment.

[0017] In addition, the purpose is to provide an AI-based real-time process control system for multi-type dispensing that estimates the virtual viscosity of the fluid in real time during the dispensing process and performs preventive parameter correction before defects occur.

[0018] Although the problem to be solved by the present invention has been described in detail, it should be noted that in addition to the aforementioned purpose, incidental purposes accompanying the achievement of the aforementioned purpose are also included within the scope of the present invention, and that achieving the aforementioned purpose partially is not necessarily required to achieve all purposes. means of solving the problem

[0020] In a preferred embodiment of the present invention, to solve the above-mentioned problem, the AI-based real-time process control system for multi-type dispensing comprises: a module identification unit that identifies the type of dispensing module that is exchangeably mounted on the dispensing equipment and outputs module identification information; an image vision unit that captures the coating result of a fluid coated on a substrate and generates a coating result image; an AI processing unit that includes a multi-type AI inspection model that has learned a set of process parameters, a coating result image, a good / bad status, and a defect type for each of the multi-type dispensing modules, and switches to an AI inspection model corresponding to the module identification information output from the module identification unit, and outputs a corrected set of process parameters according to a pre-learned defect type when the coating quality is determined to be defective based on the coating result image obtained from the image vision unit; and a system control unit that repeats the process of adjusting the process parameters of the dispensing equipment in real time according to the corrected set of process parameters of the AI ​​processing unit during the process.

[0021] In addition, in the above-described embodiment, the AI-based real-time process control system for multi-type dispensing further includes a viscosity sensing unit that measures changes in viscosity of the coating fluid in real time and outputs viscosity data, and the AI ​​processing unit preferably pre-learns the correlation between the fluid viscosity change rate and the fluid discharge amount, and compensates the corrected discharge amount based on the viscosity data detected by the viscosity sensing unit when correcting the process parameter set.

[0022] In addition, in the above-described embodiment, the AI-based real-time process control system for multi-type dispensing further includes a remaining amount detection unit that measures the remaining amount of fluid in real time and outputs remaining amount data, and the AI ​​processing unit pre-learns the correlation between the remaining amount of fluid and the fluid discharge amount, and when correcting the process parameter set, it is preferable to compensate the corrected discharge amount based on the remaining amount data detected by the remaining amount detection unit.

[0023] In addition, in the above-described embodiment, the AI-based real-time process control system for multi-type dispensing further includes a process environment sensing unit that detects the discharge pressure, flow rate, and temperature of the dispensing equipment during the coating process, and a virtual viscosity estimation unit that estimates the virtual viscosity of the fluid in real time from the discharge pressure, flow rate, and temperature detected by the process environment sensing unit, and the AI ​​processing unit preferably pre-learns the correlation between the discharge pressure, flow rate, and temperature data during the coating process and the virtual viscosity of the fluid, and preemptively compensates the corrected discharge amount based on the virtual viscosity of the fluid when correcting the set of process parameters.

[0024] In addition, in the above-described embodiment, it is preferable that the image vision unit includes a lighting module that illuminates a fluid applied on a substrate, an image detection module that generates a 2D or 3D image of the shape of the fluid applied on the substrate, and a coupling module that incorporates the lighting module and the image detection module to enable the image vision unit to be detachably attached as an independent module to a standard mounting interface of the dispensing equipment.

[0025] In addition, in the above-described embodiment, the pass / fail determination of the AI ​​processing unit is performed based on at least one of area, contour continuity, and position offset extracted from a 2D image of a coating result captured by the image vision unit, and at least one of height, volume, and cross-sectional shape extracted from 3D profile data, and the defect type classification of the AI ​​processing unit preferably includes at least one of bead breakage, width deviation, height deviation, volume deviation, position deviation, satellite, cove webbing, and tailing. Effects of the invention

[0027] According to the present invention, integrated quality management in a modular coating platform is possible by determining the quality of the coating result on a substrate in real time, regardless of the coating method.

[0028] In addition, since the AI ​​inspection model is automatically switched upon pump module replacement, the present invention enables immediate optimized inspection without additional settings by the operator, thereby improving production flexibility.

[0029] Furthermore, the present invention enables more accurate cause analysis and parameter correction compared to a single-sensor basis through composite feedback that fuses vision inspection data, viscosity data, and remaining amount data.

[0030] In addition, since the vision inspection module of the present invention is independently detachable, retrofitting to existing coating equipment is easy, which has the effect of reducing implementation costs.

[0031] In addition, the present invention has the effect of improving process yield by enabling pre-correction before defects occur through predictive control via real-time virtual viscosity estimation.

[0032] In addition, the present invention integrates the technologies of the applicant's existing registered patents (AI-based temperature-viscosity correction, pressure-heat-based residual quantity detection, PCP position control) to create synergy and strengthen the defensive power of the patent portfolio.

[0033] Although the effects obtained by the present invention have been described in detail, it should be noted that in addition to the effects described above, incidental effects obtained in the process of achieving the effects described above are also included within the scope of the effects of the present invention, and it is not necessary to obtain all effects, but rather it is based on the premise that at least a part of the effects described above is obtained. Brief explanation of the drawing

[0035] FIG. 1 is a block diagram showing the overall configuration of an AI-based real-time process control system for multi-type dispensing according to a preferred embodiment of the present invention. FIG. 2 is a perspective view showing a state in which a dispensing module and an image vision unit are mounted on a modular dispensing platform according to a preferred embodiment of the present invention. FIG. 3 is an example diagram of a bead coating result according to a preferred embodiment of the present invention. FIG. 4 is a detailed flowchart of a virtual viscosity estimation algorithm according to a preferred embodiment of the present invention. FIG. 5 is a detailed flowchart explaining the operation of an AI-based real-time process control system for multi-type dispensing according to a preferred embodiment of the present invention. Throughout the specification, including the drawings and detailed descriptions, identical elements and elements having the same function and / or the same technical or physical effect are given the same reference number or the same name, and the descriptions of elements and their functions shown or described in other embodiments are interchangeable or applicable to other embodiments. Furthermore, reference numbers described in one drawing may be omitted in other drawings to simplify identification of the drawings, and parts unrelated to the present invention or unrelated to the description of specific parts are omitted from the drawings, but this does not imply their absence. Also, In the drawings, the size of each component may be exaggerated or reduced for simplification and depicted differently from the actual scale. Specific details for implementing the invention

[0036] Below, embodiments of the present invention are described in detail with reference to the attached drawings so that those skilled in the art can easily implement the invention.

[0037] However, the present invention may be implemented in various different forms and is not limited to the embodiments described herein.

[0038] Throughout the specification, when it is stated that a part is "connected" to another part or that "transmission and reception of data" takes place, this includes not only cases where it is "directly connected" or "transmission and reception of data is directly performed," but also cases where it is "electrically connected" or "transmission and reception of data" takes place with another component in between.

[0039] Furthermore, when a part is said to "include" a certain component, this means that, unless specifically stated otherwise, it does not exclude other components but rather may include additional components, and it should be understood as not excluding in advance the existence or addition of one or more other features, numbers, steps, actions, components, parts, or combinations thereof.

[0041] Hereinafter, the overall configuration of an AI-based real-time process control system for multi-type dispensing according to a preferred embodiment of the present invention will be described with reference to FIGS. 1 to 5.

[0042] FIG. 1 is a block diagram showing the overall configuration of an AI-based real-time process control system for multi-type dispensing according to a preferred embodiment of the present invention.

[0043] Referring to FIG. 1, the AI-based real-time process control system for multi-type dispensing according to the present invention may be configured to include a dispensing unit (100), a module identification unit (200), an image vision unit (300), an AI processing unit (400), a system control unit (500), a viscosity detection unit (600), a remaining amount detection unit (700), and a virtual viscosity estimation unit (800).

[0044] Here, the term 'part' includes units realized by hardware, units realized by software, and units realized using both. Additionally, one unit may be realized using two or more pieces of hardware, or two or more units may be realized by one piece of hardware. Meanwhile, the term 'part' is not limited to software or hardware; the 'part' may be configured to reside in an addressable storage medium or configured to run on one or more processors. Thus, as an example, the 'part' includes components such as software components, object-oriented software components, class components, and task components, as well as processes, functions, attributes, procedures, subroutines, segments of program code, drivers, firmware, microcode, circuits, data, databases, data structures, tables, arrays, and variables.

[0045] The functions provided within the components and 'parts' can be combined into a smaller number of components and 'parts' or further separated into additional components and 'parts'.

[0046] First, the dispensing equipment (1000) is a device for applying fluid onto a substrate and may be configured as a modular dispensing platform including a module transfer unit (10), an integrated module (20), a dispensing unit (100), and an image vision unit (300), as shown in FIG. 2, for example.

[0047] The dispensing unit (100) and image vision unit (300) described later are integrated into an integrated module (20) and transferred to a target position on a substrate by a module transfer unit (10) made of an XY gantry structure to apply fluid.

[0048] The dispensing unit (100) is configured to allow various types of dispensing modules (110) to be replaced and mounted on a detachable module (120) provided as a standard mounting interface in the integrated module (20), and process parameters such as discharge pressure, flow rate, and speed are adjusted in real time by the system control unit (500) depending on the type of mounted module. Through this, the dispensing equipment (1000) can flexibly respond to various fluids and coating processes on a single platform.

[0049] Here, various types of dispensing modules (110) may include, for example, a jetting method, a progressive cavity pump (PCP) method, an auger method, a time-pressure method, but are not limited thereto.

[0050] Each dispensing module (110) has a different optimal method depending on the type of fluid, viscosity, application pattern and required precision. For example, the jetting method is suitable for high-speed, small-volume non-contact dispensing, the PCP method is advantageous for quantitative dispensing of high-viscosity fluids, the auger method is suitable for precise application of pastes, and the time-pressure method is suitable for continuous application of low-viscosity fluids. As such, various process requirements can be immediately met simply by changing the module, thereby maximizing production flexibility without replacing equipment.

[0051] The module identification unit (200) identifies the type of dispensing module (110) mounted on the detachable module (120) of the dispensing unit (100) and outputs the module identification information to the AI ​​processing unit (400). The module identification unit (200) can automatically recognize the type of currently mounted module by reading a unique identifier assigned to the dispensing module (110), and the unique identifier may be, for example, RFID, a barcode, an electrical signal, etc.

[0052] The module identification unit (200) automatically detects the replacement time of the dispensing module (110), reads the unique identifier of the module as soon as the new module is mounted on the detachable module (120), updates the module identification information, and outputs it to the AI ​​processing unit (400). This prevents misrecognition or non-recognition that may occur during the module replacement process and ensures that accurate module identification information is provided continuously without interruption of the process.

[0053] The image detection module (310) generates the shape of the fluid applied on the substrate as a 2D planar image and 3D profile data and transmits them to the AI ​​processing unit (400). The 2D image is a top view of the coating result taken from above and is used to extract feature values ​​such as the area, contour continuity, and position offset of the bead, and the 3D profile data is a cross-section of the coating result and is used to extract feature values ​​such as the height, volume, and cross-section shape of the bead.

[0054] The coating result image acquired by the image detection module (310) is divided into a 2D top view and a 3D cross-section as shown in FIG. 3. The 2D top view is used to check the continuity and width uniformity of the bead in a planar manner, while the 3D cross-section is used to check the height and cross-sectional shape of the bead in three dimensions. By utilizing the two data complementarily, defect types that are difficult to identify from a single viewpoint can be accurately detected.

[0055] Specifically, the defect types of the coating result images acquired by the image detection module (310) are classified into five types: (A) Normal, (B) Break, (C) Width Deviation, (D) Height Deviation, and (E) Position Deviation. Break is identified as a break in the bead in the 2D plan view, Width Deviation is identified as a change in the width of the bead in the 2D plan view, Height Deviation is identified as a change in the height of the bead in the 3D cross-section, and Position Deviation is identified as a position offset of the bead in both the 2D plan view and the 3D cross-section. The image data acquired in this manner is input into the AI ​​inspection model of the AI ​​processing unit (400) and used as basic data for determining defect types and correcting process parameters.

[0056] Meanwhile, the image detection module (310) is detachably coupled to the dispensing equipment (1000) as an independent module through the detachable module (330), so it can be operated by replacing it with a 2D camera module or a 3D profile measurement module depending on the characteristics of the coating process or inspection requirements. For example, the image detection module (310) is detachably coupled as an independent module to a standard mounting interface provided by the detachable module (330) of the integrated module (20), and is positioned side by side with the detachable module (120) equipped with the dispensing module (110) as shown in FIG. 2 to capture the coating result in real time. Accordingly, by replacing it with a 2D camera module or a 3D profile measurement module depending on the characteristics of the coating process or inspection requirements, vision inspection can be flexibly performed on a single platform in response to various fluids and coating patterns.

[0057] Meanwhile, the lighting module (320) is attached to the detachable module (330) together with the image detection module (310) and provides lighting for photographing the coating result of the fluid applied on the substrate. The lighting module (320) irradiates a uniform amount of light so that the image detection module (310) can accurately detect the shape, contour, position, etc. of the coating result, and is configured to adjust the lighting method and amount of light according to the type of coating fluid and the material of the substrate.

[0058] In particular, the lighting module (320) is integrally attached to the detachable module (330) together with the image detection module (310), thereby enabling the entire image vision unit (300) to be detachably attached as an independent module to the standard mounting interface of the dispensing equipment. This allows for easy replacement with a module that has a different lighting method depending on the characteristics of the coating process or inspection requirements, and enables the establishment of a flexible vision inspection environment that responds to various coating patterns and fluids. Meanwhile, in another example, the image detection module (310) and the lighting module (320) may be configured to be detachably attached to the detachable module (330) separately.

[0059] Meanwhile, the multi-type AI inspection model included in the AI ​​processing unit (400) is a model that has been pre-learned for each type of dispensing module (110), including a set of process parameters, an image of the coating result, whether it is good or bad, and a type of defect. As soon as module identification information is received from the module identification unit (200), it is automatically switched to an AI inspection model optimized for the corresponding module. At this time, the defect types learned by the AI ​​inspection model include bead break, width deviation, height deviation, position deviation, etc., and by outputting a corrected set of process parameters corresponding to each defect type, the coating quality is maintained in real time during the process.

[0060] For example, the AI ​​processing unit (400) can store at least two of the following: an AI inspection model for the jetting method, an AI inspection model for the progressive cavity pump (PCP) method, an AI inspection model for the auger method, and an AI inspection model for the time-pressure method. Each model is a deep learning model individually trained on the coating pattern and defect type unique to the corresponding coating method.

[0061] For example, an AI inspection model for the jetting method can be trained on defect types specific to the jetting method, such as satellite, tailing, and position deviation; an AI inspection model for the PCP method can be trained on defect types specific to continuous bead application, such as bead breakage, width deviation, and volume deviation; an AI inspection model for the auger method can be trained on defect types specific to the auger method, such as cob webbing and height deviation; and an AI inspection model for the time-pressure method can be trained on defect types specific to the time-pressure method, such as application amount deviation and position deviation.

[0062] The AI ​​processing unit (400) automatically selects and loads an AI inspection model corresponding to the type of the currently installed coating module from among a plurality of stored AI inspection models based on the type information of the coating module received from the module identification unit (200). Through this, an optimized AI inspection model is automatically activated simultaneously with the replacement of the coating module without separate settings by the operator.

[0063] The AI ​​processing unit (400) analyzes an image of a coating result obtained from an image vision unit (300) using a selected AI inspection model and determines the quality of the coating. The determination of quality is based on at least one of an area, contour continuity, and position offset extracted from a 2D image of a coating result captured by the image vision unit (300), and at least one of a height, volume, and cross-sectional shape extracted from 3D profile data.

[0064] If the pass / fail judgment result is defective, the AI ​​processing unit (400) classifies the type of defect. The classification of defect types includes at least one of bead breakage, width deviation, height deviation, volume deviation, positional deviation, satellite, cove webbing, and tailing, and the classified defect types are transmitted to the system control unit (500) and used to adjust operating parameters.

[0065] The system control unit (500) adjusts at least one operating parameter of the dispensing equipment according to the result of the AI ​​processing unit (400) and the type of defect.

[0066] The operating parameters adjusted by the system control unit (500) vary depending on the application method. For example, the system control unit (500) can adjust the valve opening / closing time, stroke, supply pressure, etc. in the case of the jetting method, the rotor rotation speed, rotation amount, supply pressure, etc. in the case of the PCP method, the screw rotation speed and rotation amount in the case of the auger method, and the supply pressure and discharge time in the case of the time-pressure method.

[0067] In addition, the system control unit (500) can further correct the operating parameters based on data received from the viscosity detection unit (600), the remaining amount detection unit (700), and the virtual viscosity estimation unit (800) described later.

[0068] The viscosity detection unit (600) measures changes in the viscosity of the coating fluid in real time and outputs the viscosity data to the AI ​​processing unit (400). The viscosity of the coating fluid changes due to various factors, such as changes in external temperature and changes in physical properties due to long-term use. When the viscosity changes, a phenomenon may occur where the fluid discharge amount changes even under the same pressure conditions. Accordingly, the AI ​​processing unit (400) pre-learns the correlation between the fluid viscosity change rate and the fluid discharge amount, and additionally compensates for the corrected discharge amount when correcting the process parameter set based on the viscosity data provided by the viscosity detection unit (600), thereby ensuring that a constant discharge amount is maintained throughout the coating process regardless of changes in fluid viscosity. At this time, the correlation between the fluid viscosity change rate and the fluid discharge amount and the additional compensation for the corrected discharge amount can be learned based on registered patent No. 10-2466306 filed by the applicant of the present invention.

[0069] In this way, by learning the correlation between the fluid viscosity change rate and the fluid discharge amount and the additional compensation of the corrected discharge amount in the AI ​​processing unit (400) based on registered patent No. 10-2466306, etc., the change in discharge amount according to the viscosity deviation by type or batch of liquid can be finely compensated even under the same external temperature conditions. Through this, the AI ​​processing unit (400) can determine an optimal discharge amount control value by comprehensively considering external temperature information and viscosity measurement values, and can maintain consistent discharge quality even for liquids with various viscosity characteristics (silicone, epoxy, resin, etc.).

[0070] Meanwhile, the remaining amount detection unit (700) measures the remaining amount of fluid in the syringe in real time and outputs the remaining amount data to the AI ​​processing unit (400). As the application process progresses, the fluid in the syringe is gradually consumed, causing the remaining amount to decrease. As the remaining amount decreases, the empty space inside the syringe expands, which may cause a phenomenon where the fluid discharge amount changes even under the same pressure conditions. Accordingly, the AI ​​processing unit (400) pre-learns the correlation between the remaining amount of fluid and the fluid discharge amount, and additionally compensates for the corrected discharge amount when correcting the process parameter set based on the remaining amount data provided by the remaining amount detection unit (700), thereby ensuring that a constant discharge amount is maintained throughout the application process. At this time, the correlation between the remaining amount of fluid and the fluid discharge amount and the additional compensation for the corrected discharge amount can be learned by the AI ​​processing unit (400) based on registered patents No. 10-2330271 and No. 10-2681795, etc., filed by the applicant.

[0071] In this way, by learning the correlation between the remaining fluid amount and the fluid discharge amount and the additional compensation of the corrected discharge amount in the AI ​​processing unit (400) based on registered patent No. 10-2330271 and registered patent No. 10-2681795, etc., the change in discharge amount caused by the expansion of the empty space due to the decrease in the remaining fluid amount in the syringe can be finely compensated even under the same pressure conditions. Through this, the AI ​​processing unit (400) can determine an optimal discharge amount control value by comprehensively considering the remaining fluid amount information and the discharge amount measurement value, and can maintain consistent discharge quality for various fluids (silicone, epoxy, resin, etc.) throughout the coating process in which the remaining amount in the syringe changes.

[0072] The virtual viscosity estimation unit (800) estimates the virtual viscosity of the fluid based on discharge pressure, flow rate, and temperature data collected in real time from the process environment sensing unit during the coating process and outputs it to the AI ​​processing unit (400). Since a physical property value corresponding to the viscosity of the fluid can be indirectly calculated from physical quantities measurable during the process without direct viscosity measurement, additional compensation can be provided preventively by supplementing the direct measurement method of the viscosity sensing unit (600).

[0073] The AI ​​processing unit (400) learns in advance the correlation between the virtual viscosity of the fluid and the discharge pressure, flow rate, and temperature data during the coating process based on the contents disclosed in Registered Patent No. 10-2798555 filed by the present applicant, and preemptively compensates the corrected discharge amount based on the estimated virtual viscosity when correcting the process parameter set, thereby enabling preemptive response to the discharge amount deviation caused by viscosity change before it leads to coating defects.

[0074] For example, referring to FIG. 4, discharge pressure, flow rate, and temperature data are collected in real time from a process environment sensing unit (not shown), and preprocessing such as noise removal and normalization is performed (S110).

[0075] The measurement data preprocessed in step S110 is applied to the Hagen-Poiseuille equation to formulate the physical relationship between fluid viscosity, flow rate, and pressure, and the initial estimate of the virtual viscosity is inversely calculated (S120).

[0076] Based on the initial viscosity estimate calculated in S120, the Newton-Raphson numerical analysis method is applied to iteratively update the virtual viscosity estimate, thereby increasing the convergence precision (S130).

[0077] The virtual viscosity estimate updated through Newton-Raphson calculation is transmitted to the corresponding AI inspection model of the AI ​​processing unit (400) to switch the AI ​​inspection module so that correction processing optimized for the currently installed dispensing module can be performed (S140).

[0078] It is determined whether the estimated virtual viscosity value satisfies the convergence condition (S150). If it does not converge, the process returns to step S120 and repeats the subsequent steps. If it converges, the final virtual viscosity estimate value confirmed to converge is output to the AI ​​processing unit (400) to be reflected in the preventive compensation of the corrected discharge amount of the process parameter set (S160).

[0079] Hereinafter, to facilitate understanding of the present invention, the operation of an AI-based real-time process control system for multi-type dispensing according to a preferred embodiment of the invention will be described with reference to FIG. 5.

[0080] FIG. 5 is a detailed flowchart illustrating the operation of an AI-based real-time process control system for multi-type dispensing according to a preferred embodiment of the present invention.

[0081] In the following description, to avoid redundancy, details specifically explained in the above description are omitted.

[0082] First, when the system is started, initial parameters required for the coating process, such as discharge pressure, flow rate, and speed, are set (S210). The initial parameter setting is a process in which a set of predefined process parameters is set in the system control unit (500) according to the type of substrate to be coated, fluid characteristics, target coating pattern, etc., and is subsequently used as a reference value throughout the process.

[0083] Next, a dispensing module (110) suitable for the application process is mounted on the detachable module (120) of the dispensing equipment (1000) (S220). The dispensing module (110) is selected and mounted according to process requirements from among various types of modules, such as the jetting method, progressive cavity pump (PCP) method, auger method, and time-pressure method.

[0084] When a dispensing module (110) is installed, the module identification unit (200) reads a unique identifier (RFID, barcode, electrical signal, etc.) assigned to the module to automatically identify the type of currently installed module and outputs the module identification information to the AI ​​processing unit (400) (S230).

[0085] The AI ​​processing unit (400) automatically switches to an AI inspection model corresponding to the received module identification information (S240). Through this, an AI inspection model optimized for the dispensing method and defect type of the currently installed dispensing module (110) is activated, and the switch is made immediately simultaneously with the module replacement without any separate settings by the operator.

[0086] Subsequently, under the control of the system control unit (500), the dispensing equipment (1000) applies fluid onto the substrate (S250). During the application process, the image vision unit (300) captures the application result in real time to generate 2D planar images and 3D profile data, and transmits them to the AI ​​processing unit (400). In addition, the viscosity detection unit (600), the remaining amount detection unit (700), and the virtual viscosity estimation unit (800) measure or estimate the viscosity, remaining amount, and virtual viscosity of the fluid in real time during the application process, respectively, and output them to the AI ​​processing unit (400).

[0087] The AI ​​processing unit (400) analyzes the coating result image using the converted AI inspection model and determines the quality of the coating by comprehensively considering viscosity data, residual amount data, and virtual viscosity estimate values, and determines whether there is a defect or a defect is expected (S260). If there is a defect or a defect is expected, the AI ​​processing unit (400) outputs a set of corrected process parameters according to the defect type, and the system control unit (500) corrects process parameters such as discharge pressure, flow rate, and speed in real time based on this (S270). The corrected parameters are immediately reflected in the coating process of S250 so that the coating quality can be improved thereafter.

[0088] Until the work is completed (S280), the steps S250 to S270 described above are repeated, and parameters are corrected in real time to improve the coating quality during the coating process.

[0089] Therefore, according to the present invention, the final coating quality can be improved well in real time.

[0090] The foregoing description of the present invention is for illustrative purposes only, and those skilled in the art will understand that other specific forms can be easily modified without altering the technical spirit or essential features of the present invention. Therefore, the embodiments described above should be understood as illustrative in all respects and not restrictive. For example, each component described as a single unit may be implemented in a distributed manner, and components described as distributed may likewise be implemented in a combined form.

[0091] The scope of the present invention is defined by the claims set forth below rather than by the detailed description above, and all modifications or variations derived from the meaning and scope of the claims and equivalent concepts thereof should be interpreted as being included within the scope of the present invention. Explanation of the symbols

[0093] 10: Module transfer section 20: Integrated Module 100: Dispensing Department 110: Dispensing Module 120: Detachable Module 200: Module Identifier 300: Image Vision Unit 310: Image detection module 320: Lighting Module 330: Detachable Module 400: AI Processing Unit 500: System Control Unit 600: Viscosity sensing unit 700: Remaining amount detection unit 800: Virtual viscosity estimation unit 1000: Dispensing equipment

Claims

Claim 1 As an AI-based real-time process control system for multi-type dispensing, the system comprises: a module identification unit that identifies the type of dispensing module interchangeably mounted on a dispensing equipment and outputs module identification information; an image vision unit that captures the coating result of a fluid applied on a substrate and generates a coating result image; a viscosity detection unit that measures changes in the viscosity of the applied fluid in real time and outputs viscosity data; a residual amount detection unit that measures the remaining amount of fluid in real time and outputs residual amount data; a multi-type AI inspection model that has learned process parameter sets, coating result images, pass / fail status, and defect types for each of the multi-type dispensing modules, and switches to an AI inspection model corresponding to the module identification information output from the module identification unit; and when the coating quality is determined to be defective based on the coating result image acquired from the image vision unit, outputs a corrected process parameter set according to a pre-learned defect type, and pre-learns the correlation between the fluid viscosity change rate and the fluid discharge amount, and compensates the corrected discharge amount based on the viscosity data detected by the viscosity detection unit when correcting the process parameter set; and according to the corrected process parameter set of the AI ​​processing unit, the AI-based real-time process control system for multi-type dispensing, comprising a system control unit that repeats the process of adjusting process parameters of dispensing equipment in real time during the process. Claim 2 delete Claim 3 delete Claim 4 In claim 1, the AI-based real-time process control system for multi-type dispensing further comprises: a process environment sensing unit that detects the discharge pressure, flow rate, and temperature of the dispensing equipment during the coating process; and a virtual viscosity estimation unit that estimates the virtual viscosity of the fluid in real time from the discharge pressure, flow rate, and temperature detected by the process environment sensing unit, wherein the AI ​​processing unit pre-learns the correlation with the virtual viscosity of the fluid from the discharge pressure, flow rate, and temperature data during the coating process, and preemptively compensates the corrected discharge amount based on the virtual viscosity of the fluid when correcting the set of process parameters. Claim 5 In claim 1, the image vision unit comprises: a lighting module that illuminates a fluid applied on a substrate; an image detection module that generates a 2D or 3D image of the shape of the fluid applied on the substrate; and a detachable module that attaches and detaches the lighting module and the image detection module as independent modules to the dispensing equipment, an AI-based real-time process control system for multi-type dispensing. Claim 6 In claim 5, the pass / fail determination of the AI ​​processing unit is performed based on at least one of area, contour continuity, and position offset extracted from a 2D image of a coating result captured by the image vision unit, and at least one of height, volume, and cross-sectional shape extracted from 3D profile data, and the defect type classification of the AI ​​processing unit includes at least one of bead breakage, width deviation, height deviation, volume deviation, position deviation, satellite, cove webbing, and tailing, an AI-based real-time process control system for multi-type dispensing.

Images