A method and system for purifying plate gum control

Abstract

The application belongs to the technical field of industrial control and relates to a purification plate glue brushing control method and system. The method comprises the following steps: obtaining an inlet plate state and corresponding nozzle channel execution state, conveying state and post-spraying pattern reading state; performing time delay compensation according to the conveying state to obtain a plate running coordinate and build a coordinate state chain; reading an effective source set and an effective state amount, generating a robust baseline residual in a neighboring stable window; generating a cross-source conflict degree, obtaining a quality gate confidence and determining a gate state; generating a control vector and outputting a control amount when the control condition is met, and writing a review-confirmed effective control result into a batch process archive. The technical scheme of the application realizes closed-loop control of the continuous glue brushing process of the purification plate through multi-source state alignment, robust residual analysis, cross-source conflict gating and result rewriting, can improve control basis consistency and control amount adjustment stability, and meets the reviewable regulation and control demand of continuous batch glue brushing.

Description

Technical Field

[0001] This application belongs to the field of industrial control technology, and specifically relates to a method and system for controlling the application of adhesive to a cleanroom panel. Background Technology

[0002] In existing technologies, the control of adhesive application to cleanroom panels typically relies on inlet detection, conveyor encoders, nozzle execution feedback, and post-spray graphic reading checks. This is achieved by triggering the nozzle channel according to the conveying position, adjusting the spraying action according to preset process parameters, and performing post-spray graphic reading checks to meet the consistency control requirements during continuous conveying of cleanroom panels. However, existing cleanroom panel adhesive application control methods have some significant shortcomings in terms of multi-source state coordination and stable feedback of control results.

[0003] In practical applications, the cleanroom panels travel continuously on the conveyor line, and inlet identification, nozzle response, conveyor speed changes, and post-spraying graphic readings are often distributed at different locations and time points. Although existing control methods can perform spraying control and result checking based on encoder position, nozzle trigger signals, or post-spraying graphic reading results, the temporal correspondence of various states during the panel's movement is easily affected by conveyor fluctuations, nozzle response lags, post-spraying graphic reading delays, and boundary area interference. Furthermore, when fluctuations in the state of the previous batch, local graphic reading anomalies, or feedback from a single source are directly used to adjust control quantities, the stability of anomaly attribution during the glue application control process is weak, and the consistency of control quantity adjustments is low, impacting the traceability and controllability of glue application results for consecutive batches of cleanroom panels.

[0004] Therefore, it is evident that existing technologies often suffer from problems such as weak coordination stability among multiple sources and low consistency in control quantity adjustment during continuous adhesive application to the cleanroom panel. These are the shortcomings of existing technologies.

[0005] In view of this, it is necessary to provide a method and system for controlling the application of adhesive to cleanroom panels in order to solve the above-mentioned defects in the prior art. Summary of the Invention

[0006] The purpose of this application is to provide a method and system for controlling the application of adhesive to cleanroom panels, in order to address the shortcomings of the prior art, such as weak stability of multi-source state coordination and low consistency of control quantity adjustment during continuous adhesive application control of cleanroom panels.

[0007] To achieve the above objectives, this application provides the following technical solution: Firstly, this application provides a method for controlling the application of adhesive to a cleanroom panel, comprising: Step S1: Obtain the inlet plate status, and the nozzle channel execution status, conveying status, and post-spraying graphic reading status corresponding to the inlet plate status; Step S2: Perform time delay compensation on the inlet plate state, nozzle channel execution state, and post-spray graphic reading state according to the conveying state to obtain the plate travel coordinates, and construct a coordinate state chain according to the plate travel coordinates; Step S3: Read the effective source set and effective state variables from the coordinate state chain, and generate robust baseline residuals within the nearest stable window based on the effective state variables; Step S4: Generate cross-source conflict degree based on robust baseline residuals and coordinate state chain; obtain quality gating confidence based on cross-source conflict degree, effective source set, current coordinate interval and robust baseline residuals; and determine gating state based on quality gating confidence and current coordinate interval. Step S5: When the quality gating confidence meets the control conditions, the cross-source conflict degree does not exceed the conflict limit, and the gating state is an allowable control state, generate a control vector based on the robust baseline residual, the gating state, and the quality gating confidence, and output the control quantity according to the control vector. Write the verified and confirmed effective control results into the batch process file.

[0008] By adopting the above technical solution, the entry identification, nozzle execution, conveying feedback and post-brush reading during the continuous conveying process of the cleanroom board are incorporated into the same control logic. This enables the coordinated cooperation between multi-source state alignment, abnormal direction constraint, control output gating and batch result write-back. It can maintain the consistency of control basis and the continuity of control actions even when the board speed fluctuates, the nozzle response is delayed, and the post-brush image acquisition is delayed. This meets the requirements of strong multi-source state coordination stability and high consistency of control quantity adjustment during the continuous glue application process of the cleanroom board.

[0009] Specifically, the simultaneous entry of the inlet, execution, conveying, and reading states into the control process establishes a clear state foundation for subsequent judgments, reducing the impact of single-source fluctuations on the overall control direction. Time delay correction around the conveying process and the formation of the board entry diameter help unify the state attribution at different locations and sampling times, ensuring continuous connection between nozzle action, board position, and post-brush reading. A locally robust reference is formed based on the effective state, compressing the offset effects caused by boundary interference, local reflections, and short-term intermittent readings, providing a more stable basis for control judgments. Gating judgments are formed by combining abnormal direction, source validity, and current control position, constraining single-point anomalies, source conflicts, and coordinate misalignments, diverting or blocking abnormal influences before they enter the control output. Control output is formed only when the gating conditions are met, and the effective batch results are written back, which helps maintain consistency between nozzle channels, trigger phases, and conveying corrections, and provides a long-term reliable and stable basis for on-site parameter verification, promoting a verifiable and traceable closed-loop control effect for continuous batch gluing processes.

[0010] Preferably, step S2, which involves performing time delay compensation on the inlet plate state, nozzle channel execution state, and post-spraying graphic reading state according to the conveying state to obtain the plate's travel coordinates, specifically includes: The plate travel reference is generated according to the conveying status. The plate travel reference includes the time coordinate mapping relationship determined according to the conveying speed and encoder status. The inlet projection position, nozzle trigger projection position, and graphic reading projection position are obtained based on the inlet plate state, nozzle channel execution state, and post-spray graphic reading state, respectively. The response delay residual is obtained based on the coordinate deviation of the inlet projection position, the nozzle trigger projection position, and the graphic reading projection position relative to the plate travel reference. When the response delay residual satisfies the same-direction offset condition within a continuous stable window, the response delay confidence interval is updated based on the response delay residual. Based on the response delay confidence interval, delay compensation is performed on the inlet plate state, nozzle channel execution state, and post-spray graphic reading state to obtain the plate travel coordinates; When the response delay confidence interval exceeds the allowable boundary, the corresponding source is written into the source sequence to be reviewed, and the corresponding source does not participate in the control judgment of the current coordinate interval.

[0011] Based on the above scheme, a correction relationship between the plate inlet diameter and time delay offset is established around the conveying process. This enables inlet identification, nozzle triggering, and post-brush reading to be aligned under the same travel reference. This reduces the superimposed impact of conveying speed fluctuations, nozzle response lag, and acquisition delay on control judgment. Furthermore, it forms a verification constraint when the source offset exceeds the boundary, thereby improving the stability of the control basis within the current coordinate range and the controllability of the source participation boundary.

[0012] Preferably, step S2, which involves constructing a coordinate state chain based on the plate's travel coordinates, specifically includes: The inlet compensation coordinates, nozzle trigger compensation coordinates, and graphic reading compensation coordinates are obtained based on the plate travel coordinates and the response delay confidence interval, respectively. A coordinate assignment window is generated based on the coordinate overlap relationship between the inlet compensation coordinates, the nozzle trigger compensation coordinates, and the graphic reading compensation coordinates. The following markers are determined according to the coordinate attribution window: inlet status integrity marker, boundary positioning reliability marker, conveying speed stability marker, nozzle channel feedback integrity marker, post-spray graphic reading continuity marker, and coordinate attribution consistency marker. When there are states within the same coordinate attribution window that do not meet the source validity condition, write the states that do not meet the source validity condition into the isolated source sequence; A coordinate state chain is generated based on the plate's travel coordinates, the coordinate ownership window, and the states within the coordinate ownership window that have not been written with an isolated source sequence.

[0013] In the above scheme, the plate movement coordinates are further transformed into coordinate attribution criteria after multi-source compensation, so that the states of the entry side, execution side and reading side can form an effective association around the same plate position. Abnormal states are filtered by source validity constraints, which can improve the clarity and availability of state attribution in the coordinate state chain, reduce the risk of invalid sources entering subsequent residual analysis and conflict judgment, and enhance the reliability of the state chain in continuous transportation scenarios.

[0014] Preferably, step S3, which involves generating robust baseline residuals within a nearby stable window based on effective state variables, specifically includes: Determine the source validity distribution of the current coordinate interval based on the valid state quantities; Candidate neighbor windows are determined from the coordinate state chain based on the source validity distribution, and the inlet transition region, tail convergence region, boundary avoidance region and coordinate intervals where the source validity does not meet the conditions are removed from the candidate neighbor windows to obtain the neighbor stable window; Read the continuous data of the post-spray graphic within the adjacent stable window to generate a set of benchmark samples; Local baseline values ​​and scale quantities are obtained from the baseline sample set, and robust baseline residuals are generated by reading continuous quantities, local baseline values, and scale quantities from the post-spraying graphic corresponding to the current coordinate interval.

[0015] Based on the above processing, a nearby stable reference is established around the effective state, so that the baseline generation avoids the abnormal areas of the entrance, tail end, boundary and source. This can reduce the impact of transition state and local disturbance on the comparison basis, and make the changes in the brushed pattern in a more stable state range, thereby enhancing the ability of the robust baseline residual to characterize the actual brushing state deviation.

[0016] Preferably, the steps of obtaining local baseline values ​​and scale quantities based on the baseline sample set, and generating robust baseline residuals by reading continuous quantities, local baseline values, and scale quantities from the post-jetting pattern corresponding to the current coordinate interval, specifically include: The local benchmark value is obtained based on the median of the benchmark sample set; The scale quantity is obtained by reading the median absolute deviation or interquartile range of the continuous quantity relative to the local reference value from the sprayed graphic in the reference sample set. Based on the difference between the continuous quantity and the local reference value read from the post-spraying graphic corresponding to the current coordinate interval, the scale quantity is normalized to obtain the robust baseline residual. When the scale quantity exceeds the allowable boundary of the stability window, the neighboring stability windows are marked as unusable windows, and a new neighboring stability window is selected based on the source validity distribution. The robust baseline residuals are updated based on the newly selected nearest stable window.

[0017] Based on the above scheme, robust statistical methods are used to form local references and scale constraints, and neighboring stable windows are reselected when scale anomalies occur. This makes it less likely for short-term discontinuous patterns, local reflections, or isolated anomalous samples to dominate the residual results, thereby improving the robust baseline residuals' resistance to disturbances and cross-batch comparability.

[0018] Preferably, step S4, which generates the cross-source conflict degree based on the robust baseline residuals and coordinate state chain, specifically includes: Based on the coordinate state chain, identify the abnormal amount of post-spray graphic reading, nozzle trigger confirmation deviation, response current deviation, speed fluctuation, encoder jump and position residual out-of-window status within the current coordinate range; The abnormal reading amount of the post-jetting pattern that is consistent with the direction of the robust baseline residual and satisfies the residual magnitude boundary is classified as the reading-side conflict amount. The nozzle trigger confirmation deviation and response current deviation are classified into the execution-side conflict quantity; Speed ​​fluctuations and encoder jumps are classified as conveyor-side conflict quantities. The position residual window states from different sources within the same current coordinate interval are classified into coordinate attribution conflict quantities. A quaternary conflict quantity is generated based on the conflict quantity on the read side, the conflict quantity on the execution side, the conflict quantity on the transmission side, and the conflict quantity on the coordinate attribution side. The cross-source conflict degree is obtained by weighting and summing the conflict quantities according to the source weight and coordinate attribution weight of each conflict quantity in the quaternary conflict quantity.

[0019] In the above scheme, a multi-source anomaly attribution method is formed around the coordinate state chain, so that graphic reading deviation, execution feedback deviation, transmission fluctuation and position attribution anomaly are included in a unified conflict expression. This can avoid a single anomaly source directly dominating the control judgment and make the cross-source conflict degree more accurately reflect the comprehensive stability of the glue application control state within the current coordinate range.

[0020] Preferably, step S4, which involves obtaining the quality-gated confidence level based on the cross-source conflict degree, the valid source set, the current coordinate interval, and the robust baseline residuals, specifically includes: Update the source weights based on the number of times each source is available and isolated within the adjacent coordinate allocation window in the valid source set; Update the coordinate assignment weights based on the position residual window persistence state of the coordinate assignment conflict quantity within the current coordinate interval. A state deviation deduction is generated based on the deviation between the robust baseline residual and the residual trigger boundary corresponding to the current coordinate interval. When the conflict quantity on the read side meets the conflict condition, but the conflict quantity on the execution side, the conflict quantity on the transmission side, and the conflict quantity on the coordinate attribution do not meet the conflict condition, the source weight corresponding to the conflict quantity on the read side is reduced, and the source corresponding to the conflict quantity on the read side is written into the review waiting source sequence. When the conflict quantity on the read side and the conflict quantity on the execution side meet the same-direction conflict condition and the conflict quantity on the coordinate attribution does not meet the conflict condition, the current coordinate interval is marked as the execution-side related conflict interval, and the quality gating confidence is obtained based on the cross-source conflict degree, the set of valid sources, the sequence of sources waiting for review, and the state deviation deduction amount corresponding to the execution-side related conflict interval.

[0021] Based on the aforementioned conflict expression logic, the availability of the source, the persistent status of the location attribution, and the residual deviation status are all transformed into gating confidence criteria. This allows single-source anomalies on the reading side to enter the review waiting phase, and related anomalies on the execution side to form independent conflict intervals. This strengthens the ability of quality gating confidence to distinguish differences in anomaly sources and improves the stability of risk constraints before control output.

[0022] Preferably, step S5, which generates a control vector based on the robust baseline residuals, gating state, and quality gating confidence, and outputs the control quantity according to the control vector, specifically includes: The current control region is determined based on the current coordinate interval, and the control vector change boundary and hysteresis condition are determined based on the current control region. The current control region includes the inlet transition region, the stable control region, the boundary avoidance region, and the tail convergence region. When the gated state is the allowed control state, the residual correction component is obtained based on the robust baseline residual and the quality gate confidence, and the control vector to be executed is generated based on the previous control vector and the residual correction component. The control vector to be executed is limited according to the control vector change boundary and hysteresis condition to obtain the control vector. The control quantity is then output according to the control vector. The control vector and the current coordinate interval are bound to generate the control result. The control quantity includes nozzle channel selection, injection trigger phase, pulse width, atomization intensity, channel shielding, delivery speed correction and execution channel gain adjustment.

[0023] Based on the above processing, the current control region, gating state, and residual correction relationship are incorporated into the control output constraints, so that the control vector is formed under the allowed control state and is restricted by the changing boundary and hysteresis conditions. This can reduce control jumps caused by boundary regions and short-term residual fluctuations, and enhance the consistency of action connection between nozzle channel, trigger phase, and delivery correction.

[0024] Preferably, step S5, which involves writing the verified and confirmed control results into the batch process file, specifically includes: The verification object is generated based on the control vector and the current coordinate range in the control results, and the verification order is determined according to the cross-source conflict degree. The verification order includes post-spraying graphic reading verification, nozzle channel feedback verification, conveying speed verification, and coordinate attribution verification. The review confirmation results are generated based on the review object and review order. The review confirmation results include review confirmation valid, review source unknown, and review confirmation invalid. When the verification result is valid, the control vector, current coordinate range, quality gate confidence level and verification result will be written into the batch process file. When the review confirmation result is that the source of the review is unknown, the control result will be written into the candidate event sequence; When the verification result is invalid or the control result causes control oscillation, the current coordinate interval is written to the disabled sample index. The data corresponding to the disabled sample index will not participate in the update of the initial control value of the next batch. Control oscillation includes the state in which the control vector direction of the same current coordinate interval in adjacent batches alternates and the amplitude of the control vector exceeds the oscillation boundary.

[0025] In the above scheme, a review order and result flow boundary are established around the control results, so that valid results enter the batch process file, results of unknown origin enter the candidate event sequence, and abnormal or oscillating results enter the disabled sample index. This can prevent abnormal samples from participating in the update of the initial control value of the next batch and enhance the traceability and closed-loop stability of continuous batch glue application control.

[0026] Secondly, this application also provides a cleanroom panel adhesive application control system, comprising: The status acquisition unit is used to acquire the status of the inlet plate, as well as the nozzle channel execution status, conveying status, and post-spraying graphic reading status corresponding to the status of the inlet plate. The coordinate construction unit is used to perform time delay compensation on the inlet plate state, nozzle channel execution state and post-spraying graphic reading state according to the conveying state, to obtain the plate travel coordinates, and to construct a coordinate state chain according to the plate travel coordinates. The residual generation unit is used to read the effective source set and effective state quantities from the coordinate state chain, and generate robust baseline residuals based on the effective state quantities within the adjacent stability window. The gating determination unit is used to generate cross-source conflict degree based on robust baseline residuals and coordinate state chain, obtain quality gating confidence based on cross-source conflict degree, effective source set, current coordinate interval and robust baseline residuals, and determine gating state based on quality gating confidence and current coordinate interval; The control output unit is used to generate a control vector based on the robust baseline residual, the gate state, and the quality gate confidence when the quality gate confidence meets the control conditions, the cross-source conflict degree does not exceed the conflict limit, and the gate state is an allowable control state. It then outputs the control quantity according to the control vector and writes the verified and confirmed effective control results into the batch process file.

[0027] As can be seen from the above technical solutions, this application has the following advantages: The present application provides a method and system for controlling the application of adhesive to a cleanroom panel. By incorporating inlet identification, nozzle execution, conveying feedback, and post-application reading during the continuous conveying of the cleanroom panel into the same control logic, it achieves coordinated cooperation between multi-source state alignment, abnormal direction constraints, control output gating, and batch result write-back. This enables the consistency of control basis and the continuity of control actions even when there are fluctuations in the panel's travel speed, lag in nozzle response, and delay in post-application image acquisition. It meets the requirements of strong multi-source state coordination stability and high consistency of control quantity adjustment during the continuous application of adhesive to the cleanroom panel. Attached Figure Description

[0028] To more clearly illustrate the technical solutions of this application, the accompanying drawings used in the embodiments are briefly described below. The following drawings only show some embodiments of this application; those skilled in the art can obtain other drawings based on these drawings without creative effort.

[0029] Figure 1 This is a flowchart of a method for controlling the application of adhesive to a cleanroom panel, as provided in this application; Figure 2 This is a schematic diagram of a cleanroom panel adhesive application control system provided in this application.

[0030] The system comprises: 1. State acquisition unit; 2. Coordinate construction unit; 3. Residual generation unit; 4. Gating determination unit; and 5. Control output unit. Detailed Implementation

[0031] The embodiments of this application are described below with reference to the accompanying drawings. It should be understood that the following embodiments are only used to illustrate the technical solutions of this application and are not intended to limit the scope of protection of this application. Any adjustments, equivalent substitutions, improvements or other optional implementation methods made by those skilled in the art to the embodiments without departing from the concept and scope of protection of this application should fall within the scope of protection of this application.

[0032] It should be noted that in the description of this application, the terms "comprising," "including," "having," and their synonyms are used to indicate the presence of the described features, structures, steps, operations, elements, components, or combinations thereof, but do not exclude the presence of other features, structures, steps, operations, elements, components, or combinations thereof.

[0033] It should be noted in advance that, in order to facilitate understanding of the technical solutions of the embodiments of this application, some terms and related technologies involved in the embodiments of this application will be briefly explained below: 1. Encoder: A detection device used to convert the position, speed or angular displacement of mechanical motion into electrical signals or digital pulse signals. It is often used in conveyor lines, servo control and automation equipment to obtain information on motion distance, speed or position change, and to provide basic feedback data for motion control and position synchronization.

[0034] 2. Median absolute deviation: A robust statistical indicator used to measure the dispersion of data. It is usually obtained by taking the median after taking the absolute deviation between each data value and the median. Compared with mean and variance indicators, it is not sensitive to isolated outliers and is suitable for data analysis scenarios with local mutations, noise or outlier samples.

[0035] 3. Interquartile Range: A statistical indicator describing the dispersion of data, usually the difference between the third quartile and the first quartile. It is used to reflect the distribution range of the middle part of the data. Compared with the range, it is less affected by extreme data such as the maximum or minimum value. It is often used for robust statistics and abnormal fluctuation analysis.

[0036] To address the issues of weak multi-source state coordination stability and low consistency of control quantity adjustment during continuous glue application control of cleanroom panels, the on-site glue application process, under the combined effects of transport fluctuations, nozzle response lag, and post-brush reading delays, struggles to meet the practical requirements of stable control basis and verifiable control results for continuous batch glue application. This application discloses a cleanroom panel glue application control method and system. By establishing a collaborative processing approach that integrates multi-source state alignment, local robust reference, cross-source conflict gating, and batch result write-back for the panel's movement, it enhances the temporal consistency of glue application control basis, the ability to constrain abnormal directions, and the continuity of control output. This improves the state coordination level, control quantity adjustment consistency, and batch traceability reliability during continuous cleanroom panel glue application.

[0037] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0038] like Figure 1 As shown in this embodiment, a method for controlling the application of adhesive to a cleanroom panel includes: Step S1: Obtain the inlet plate status, and the nozzle channel execution status, conveying status, and post-spraying graphic reading status corresponding to the inlet plate status; Step S2: Perform time delay compensation on the inlet plate state, nozzle channel execution state, and post-spray graphic reading state according to the conveying state to obtain the plate travel coordinates, and construct a coordinate state chain according to the plate travel coordinates; Step S3: Read the effective source set and effective state variables from the coordinate state chain, and generate robust baseline residuals within the nearest stable window based on the effective state variables; Step S4: Generate cross-source conflict degree based on robust baseline residuals and coordinate state chain; obtain quality gating confidence based on cross-source conflict degree, effective source set, current coordinate interval and robust baseline residuals; and determine gating state based on quality gating confidence and current coordinate interval. Step S5: When the quality gating confidence meets the control conditions, the cross-source conflict degree does not exceed the conflict limit, and the gating state is an allowable control state, generate a control vector based on the robust baseline residual, the gating state, and the quality gating confidence, and output the control quantity according to the control vector. Write the verified and confirmed effective control results into the batch process file.

[0039] This embodiment introduces a state integration method from the inlet side, execution side, conveying side, and reading side, ensuring that the continuous purification plate has a clear source, location-related, and process-corresponding state basis when entering the control process. This reduces the influence of a single detection result on the direction of glue application control. The conveying process organizes the states from different sources over time and establishes a state continuity relationship along the plate's travel dimension, helping to reduce state misalignment caused by differences in acquisition location and response delays, resulting in a more continuous connection between nozzle action, plate position, and reading results. Establishing a nearby stable reference around the effective state limits the disturbance of the judgment benchmark by transition zones, boundary zones, and short-term fluctuations, allowing for a more stable comparison of post-brush pattern changes. Based on the principle of combining residual direction and state chain relationship to form cross-source conflict expression, and further forming gating judgment, it is helpful to distinguish between single reading anomaly, execution feedback anomaly, and coordinate attribution anomaly, so that the impact of anomaly is constrained before control output; after the gating condition is met, control output is formed and valid results are written back, which can maintain the correspondence between control action and batch record, and facilitate subsequent review of control results and traceability of control basis; overall, through the continuous cooperation between state input, coordinate inheritance, robust reference, conflict gating and result writing back, the cleanroom plate gluing control can maintain a more stable collaborative control capability and engineering adaptability in the field environment where there are changes in conveying speed, nozzle response fluctuations and graphic reading disturbances.

[0040] The above steps will be specifically described below based on the embodiments of this application.

[0041] In step S1, the core task is to integrate the states from the inlet side, execution side, conveying side, and reading side into the same sampling aperture, forming a data foundation with clear sources, specific timing, and measurable quality, and outputting a multi-source state package that can be incorporated into coordinate compensation. In this embodiment, the inlet reading window is set near the inlet baseline. The inlet side reads the plate number, plate type identification, leading edge arrival mark, and edge positioning result; the execution side reads the nozzle channel trigger confirmation, response current, and channel shielding status; the conveying side reads the encoder pulse, conveying speed, and speed fluctuation; and the reading side reads the post-jet pattern reading status formed by line scanning.

[0042] Based on this, the inlet plate status and the corresponding nozzle channel execution status, conveying status, and post-spraying graphic reading status can be obtained. During this process, each status is accompanied by a source number, sampling time, installation reference position, and original quality mark. The inlet plate status not only records whether the plate has arrived, but also records the leading edge coordinates of the current plate at the inlet reference line, the reliability of the left and right boundaries, and the batch identifier, enabling subsequent determination of the attribution relationship of the same plate in different sources. The nozzle channel execution status is derived from the spray trigger drive record and nozzle channel feedback record, including the trigger command issuance time, trigger confirmation return time, response current peak value, response current stable segment, and channel shielding status. The conveying status is derived from the encoder pulse and speed estimation process, including the current conveying speed, encoder jump mark, and speed stability mark. The post-spraying graphic reading status is derived from the continuous graphic features within the reading window, including continuous quantities along the plate's travel direction, missing line segment marks, and local strong reflection marks.

[0043] Furthermore, to avoid invalid samples being mixed into control decisions during subsequent calculations, the source states at the same sampling time can be organized into a multi-source state packet; for example, the first... Multi-source state packets at each sampling time It can be represented as: ,in, Indicates the first Each state source is the original acquisition timestamp or feedback arrival time at the corresponding sampling moment. This indicates that the state originates from the original projected position obtained based on the installation reference. The field representing the original state from which this state originated. This indicates the initial screening value of the source quality. Indicates the first The set of state sources that can be read at each sampling time. This state packet serves as the input to step S2, enabling delay compensation to simultaneously access the original projection position, the original acquisition timestamp or feedback arrival time, and source quality information.

[0044] In this process, the initial screening value of the source quality It can be generated based on positional abrupt changes and original time abrupt changes between adjacent homologous state records, avoiding situations where a single source already has significant jumps before entering coordinate compensation. For example, the initial screening relationship for source quality can be: ,in, and They represent the first The allowable scale of change between adjacent locations and the scale of change between adjacent original times for each state source. and These are the position stability constant and the time stability constant, respectively. For example, the position change scale corresponding to the inlet plate state can be determined according to the upper limit of the inlet positioning error; the time change scale corresponding to the nozzle channel execution state can be determined according to the maximum response jitter of the trigger confirmation; and the position change scale corresponding to the post-spray graphic reading state can be determined according to the pixel calibration error of the reading window. When the initial screening value of the source quality is lower than the on-site set lower limit, the source is still recorded, but in step S2, it can only participate in compensation as a source to be verified and does not participate in the control judgment of the current coordinate interval.

[0045] It should be noted that when the inlet plate status is missing, a temporary plate identifier can be generated by combining the spacing between the front and rear plates, plate length, and conveying speed. This temporary plate identifier only maintains the continuity of the status packet and does not allow direct triggering of channel shielding or control vector correction. When a sudden jump in response current occurs in the nozzle channel execution status but the trigger confirmation is normal, the source can be marked as a suspected conflict source on the execution side. When a partial missing line occurs in the post-spray graphic reading status, the initial screening value of the source quality of the corresponding coordinate segment can be reduced while retaining the original reading curve for reading during the sequential generation of step S4.

[0046] Thus, step S1 establishes a multi-source status packet and a preliminary screening relationship for source quality, enabling the information from the inlet side, execution side, delivery side, and reading side to have a unified sampling caliber, providing an input basis for delay compensation and coordinate attribution in step S2.

[0047] In step S2, the core task is to convert the states obtained under different installation positions and different response delays into the plate travel coordinates, retain the valid states within the same coordinate ownership window, and output a continuously traceable coordinate state chain.

[0048] In some embodiments of this application, encoder pulses and speed estimates during the conveying process are used as the primary basis for coordinate conversion. The encoder pulses provide the cumulative travel of the sheet metal, and the speed estimates provide continuous displacement compensation between sampling times. The original acquisition timestamp or feedback arrival time is used as the source. After the installation position is written into the status packet, time delay compensation can be performed on the inlet plate status, nozzle channel execution status, and post-spray graphic reading status according to the delivery status, and the plate travel coordinates can be obtained. Among them, the sampling position of the inlet plate status is close to the inlet baseline, the feedback position of the nozzle channel execution status is located in the spray trigger area, and the sampling position of the post-spray graphic reading status is located in the reading window. There are differences in installation distance between the three in physical space and in response time. Therefore, it is necessary to establish the plate travel baseline first.

[0049] Specifically, a time coordinate mapping relationship can be established based on the encoder increment, encoder resolution, average conveying speed at the sampling time, and encoder state. After triggering the inlet baseline, a material travel reference is generated according to the conveying state. The time coordinate mapping relationship uses the inlet baseline as the coordinate zero point, converts the encoder pulse increment at the sampling time into the travel distance, and uses the conveying speed within the sampling period to compensate for the displacement between discrete encoder pulses. For example, the time coordinate mapping relationship can be written as: , in, Indicates the first The reference function for the plate movement at each sampling time point This indicates the initial travel coordinates corresponding to the entrance baseline. Indicates the first Encoder count at each sampling time, This indicates the encoder count when the entry baseline is triggered. This indicates the encoder's resolution per unit distance. Indicates the first Average transport speed within each sampling time period This indicates the trigger time of the entry baseline. For example, the encoder resolution can be calibrated according to the travel distance corresponding to the pulse, and the sampling period can be set to... - When the conveying speed fluctuation exceeds the stability boundary, the sampling window can be shortened to reduce coordinate assignment error.

[0050] Based on this, projection positions can be formed according to the physical installation location and state triggering time of each source. The production line controller obtains the inlet projection position, nozzle trigger projection position, and graphic reading projection position according to the inlet plate state, nozzle channel execution state, and post-spray graphic reading state, respectively. The inlet projection position is mainly obtained from the inlet baseline trigger coordinates and edge positioning results; the nozzle trigger projection position is mainly obtained from the nozzle installation baseline, trigger confirmation return time, and nozzle channel number; and the graphic reading projection position is mainly obtained from the reading window calibration coordinates, graphic acquisition time, and conveying state. Each projection position is compared with the plate travel baseline. To convert the deviation to the same time scale, the response delay residual can be obtained from the coordinate deviation of the inlet projection position, nozzle trigger projection position, and graphic reading projection position relative to the plate travel baseline, which can be written as: ,in, Indicates the first The state source is in the first The response delay residual corresponding to each sampling time point This represents the velocity stability constant. The positive or negative direction of the response delay residual indicates whether the source is ahead or behind the plate travel reference. The larger the absolute value of the residual, the more significant the deviation between the source and the target coordinate at the current sampling time. This residual is not only used to compensate for coordinates but also serves as the source of the position residual out-of-window state in step S4.

[0051] Furthermore, a continuous stable window can be selected from coordinate segments where the conveying speed is stable, the inlet status is complete, and the post-jet graphic readings are continuous. When the response delay residual satisfies the same-direction offset condition within the continuous stable window, the response delay confidence interval is updated based on the response delay residual. The same-direction offset condition requires that the residual directions of the same source are consistent across multiple continuous stable windows, and that the median residual value exceeds the calibration noise boundary. In this case, it can be considered that the source exhibits stable response drift, and the response delay confidence interval needs to be narrowed to an interval centered on the median residual value that can cover the stable drift fluctuation range. The confidence interval update relationship can be written as: , in, Indicates the first The state source is in the first The confidence interval of the response delay after each sampling time is updated. Indicates the first The source of the state is the first From the sampling time to the first sampling time The response delay residual sequence within a continuous stable window consisting of sampling times. This represents the median value of the response delay residual sequence. This represents the median absolute deviation of the response delay residual sequence relative to the median value. This represents the confidence amplification factor. For example, the confidence amplification factor for the inlet plate state can be 1.5-2.0, for the nozzle channel execution state (due to stronger jitter from trigger confirmation response) it can be 2.0-2.8, and for the post-spray graphic reading state (affected by lighting fluctuations in the reading window) it can be 2.5-3.0. Increasing the confidence amplification factor increases the tolerance for source retention, but also increases the risk of position mismatch.

[0052] In this embodiment, the updated response delay confidence interval participates in the coordinate compensation for each source. Specifically, delay compensation can be performed on the inlet plate state, nozzle channel execution state, and post-spraying graphic reading state according to the response delay confidence interval to obtain the plate travel coordinates. Specifically, during compensation, the center of the response delay confidence interval can be taken as the online response delay estimate of the current source, and the interval width can be retained as the coordinate confidence boundary. For example, the compensated coordinates can be written as: ,in, Indicates the first The state source is in the first The compensated coordinates of the plate at each sampling time. Indicates the first The online response delay is estimated for each state source. After compensation, the plate's traveling coordinates are entered into the coordinate attribution window for judgment. If the confidence interval width of the same source continues to expand, it indicates that the source coordinates are unstable and their control contribution to the current coordinate interval needs to be limited.

[0053] Furthermore, the allowable boundary can be determined jointly by the installation calibration error, encoder resolution, and control area width. When the response delay confidence interval crosses the allowable boundary, the corresponding source is written into the source sequence to be reviewed, and this corresponding source does not participate in the control judgment of the current coordinate interval. At the same time, the source sequence to be reviewed records the source number, the direction of the boundary crossing, the number of times the boundary crossing occurred, and the most recent original state field. Subsequent reviews can use this information to decide whether to redo the source calibration or isolate the source only in the current batch.

[0054] In some embodiments of this application, after completing delay compensation and obtaining the plate's travel coordinates, a coordinate attribution window can be established around the target coordinates, and a coordinate state chain can be constructed according to the plate's travel coordinates. Specifically, the inlet compensation coordinates, nozzle trigger compensation coordinates, and image reading compensation coordinates can be obtained based on the plate's travel coordinates and the response delay confidence interval, respectively. The inlet compensation coordinates are used to determine the plate's identity and leading edge position, the nozzle trigger compensation coordinates are used to determine the actual landing point of the channel action, and the image reading compensation coordinates are used to determine the plate position corresponding to the continuous image reading amount after spraying.

[0055] In this process, the coordinate assignment window is based on the center of the current coordinate interval and the response delay confidence interval width is converted into a window tolerance. Therefore, the coordinate assignment window is generated based on the coordinate overlap relationship between the inlet compensation coordinates, nozzle trigger compensation coordinates, and graphic reading compensation coordinates, which can be represented as: ,in, Indicates the first The coordinates of the current coordinate interval belong to the window. Indicates the first The center travel coordinate of the current coordinate interval Indicates the first The allowable deviation for coordinate assignment within the current coordinate interval. The allowable deviation for coordinate assignment can be smaller in the stable control zone, and larger in the entrance transition zone and the tail convergence zone. The boundary avoidance zone is tightened according to the distance from the edge of the plate to prevent the reading state near the boundary from mistakenly entering the control judgment of adjacent channels.

[0056] Furthermore, based on the coordinate attribution window, the following markers are determined: inlet status integrity marker, boundary positioning reliability marker, conveyor speed stability marker, nozzle channel feedback integrity marker, post-spray graphic reading continuity marker, and coordinate attribution consistency marker. Specifically, the inlet status integrity marker reflects whether the plate identity and inlet trigger are complete; the boundary positioning reliability marker reflects whether the left and right boundary offsets from the center are within the positioning error range; the conveyor speed stability marker reflects whether encoder jumps and speed fluctuations are within the allowable range of the sampling window; the nozzle channel feedback integrity marker reflects whether the trigger confirmation and response current correspond; the post-spray graphic reading continuity marker reflects whether there is a break in the continuous linear scan; and the coordinate attribution consistency marker reflects whether the three compensation coordinates fall within the same coordinate attribution window.

[0057] Based on this, when a state exists within the same coordinate attribution window that does not meet the source validity conditions, that state is written into the isolated source sequence. The isolated source sequence records not only the source number but also the isolation reason, isolation start point, isolation duration window number, and recoverable conditions. Further, a coordinate state chain is generated based on the board's travel coordinates, the coordinate attribution window, and the states within the coordinate attribution window that are not written into the isolated source sequence. The coordinate state chain is arranged according to the board's travel direction, and each state node contains the current coordinate interval, the set of valid sources, the initial screening value of the source quality, the compensation coordinates, six types of markers, and the original state reference. Step S3 reads the set of valid sources and the valid state quantity from this chain.

[0058] Thus, step S2 completes the transformation from the conveying state to the plate travel reference, from the multi-source projection position to the compensation coordinate, and from the coordinate ownership window to the coordinate state chain, providing a reliable coordinate basis for the subsequent robust baseline residual generation.

[0059] In step S3, the core task is to filter out the valid states that can be compared from the coordinate state chain, form a local robust reference within the adjacent stable window, and output a robust baseline residual that can reflect the degree of deviation of the current coordinate interval.

[0060] In this embodiment, the effective source set and effective state quantities can be read from the coordinate state chain, and robust baseline residuals can be generated within the adjacent stable window based on the effective state quantities. The effective source set mainly consists of sources that have not entered the source sequence to be reviewed or the isolated source sequence. The effective state quantities mainly come from the continuous data read from the post-spraying graphic, and are combined with the inlet state integrity, conveying speed stability marker, and nozzle channel feedback integrity marker to determine their comparability. Furthermore, before an effective state quantity enters the adjacent stable window, it needs to be confirmed that the source to which the effective state quantity belongs, the coordinate window to which it belongs, and the continuous data read from the post-spraying graphic are all in a comparable state. Only state nodes that can form a proximity relationship with the current coordinate interval to the same plate material are included in the subsequent benchmark sample screening.

[0061] In some embodiments of this application, the source validity distribution of the current coordinate interval can be determined based on the valid state quantity. The source validity distribution records the available, missing, isolated, or pending verification status of each source within the current coordinate interval, and subsequent candidate neighbor windows cannot cross obvious source missing segments. Based on this, candidate neighbor windows are determined from the coordinate state chain according to the source validity distribution. The candidate neighbor windows extend forward and backward with the current coordinate interval as the center, and the extension length is determined based on the plate length, the stable control zone length, and the natural fluctuation range of the continuous amount of data read from the post-spraying graphic. After the candidate neighbor windows are determined, coordinate segments that are prone to causing baseline shifts need to be removed from the candidate neighbor windows, including the inlet transition zone, the tail end convergence zone, the boundary avoidance zone, and coordinate intervals where the source validity does not meet the conditions, finally obtaining the neighboring stable window.

[0062] Based on this, continuous data from the post-spraying image is read within a nearby stable window to generate a benchmark sample set. These continuous data from the post-spraying image can be generated based on line scan grayscale stability, continuous reading line length, local missing line ratio, and channel coverage consistency, but are uniformly used as comparable continuous data when entering robust baseline calculation. For example, the first... The set of reference samples corresponding to each current coordinate interval It can be written as: ,in, Indicates the first The continuous data of the post-spray graphic reading from a number of adjacent coordinate intervals. Indicates the first The nearest coordinate intervals are selected based on the following criteria: complete inlet status, stable delivery speed, complete nozzle channel feedback, and continuous post-spray graphic reading. The benchmark sample set is read only from the nearest stable window, without using the current coordinate interval itself, to avoid current anomalies entering the local benchmark.

[0063] Furthermore, local baseline values ​​and scale quantities are obtained based on the baseline sample set. Specifically, local baseline values ​​can be obtained from the median of the baseline sample set, thereby reducing the impact of local strong reflections or isolated missing lines on the comparison baseline. Further, scale quantities are obtained from the median absolute deviation or IQR (Interquartile Range) of the continuous readings of the post-sprayed pattern in the baseline sample set relative to the local baseline values, used to represent the natural fluctuation amplitude of adjacent stable windows. In this process, when the continuous readings of the post-sprayed pattern in adjacent stable windows are relatively concentrated, the median absolute deviation can provide a more sensitive scale; when the surface texture of the same plate type causes slow fluctuations within the stable range, the interquartile range can provide a more stable scale. For example, the first... The scale corresponding to the current coordinate interval It can be written as: , in, Represents the scale stability constant. Median absolute deviation term. Reflects the absolute fluctuation of the benchmark sample set around a local benchmark value; interquartile range term This is a method in descriptive statistics used to determine the difference between the third quartile and the first quartile, which reflects the distribution width of the middle portion of a baseline sample set; taking the larger of the two values ​​can prevent slight reading fluctuations from being amplified due to an excessively small scale. For example, the scale stability constant can be set at 0.5%-1.0% of the full-scale reading.

[0064] In some embodiments of this application, robust baseline residuals can be generated based on the continuous reading of the post-spraying pattern corresponding to the current coordinate interval, the local reference value, and the scale value. Specifically, the scale value is normalized based on the difference between the continuous reading of the post-spraying pattern and the local reference value corresponding to the current coordinate interval, so that deviations under different plate types, different reading amplitudes, and different surface textures can be included in the same judgment caliber, and robust baseline residuals are obtained. For example, the first robust baseline residuals corresponding to each current coordinate interval It can be written as: ,in, Indicates the first The continuous data of the post-spray graphic reading corresponding to the current coordinate interval. Indicates the first The local reference value corresponding to the current coordinate interval. The robust baseline residual retains the positive and negative directions. Positive deviation and negative deviation respectively indicate that the current coordinate interval has different control meanings relative to the adjacent stable window. Step S4 will continue to determine whether the residual has an execution-side correspondence based on the nozzle trigger confirmation deviation, response current deviation and delivery-side changes.

[0065] It should be noted that the adjacent stable window is not fixed. The allowable boundary of the stable window can be determined by combining historical stable readings of the same board type, the current board width, and the reading window calibration error. When the scale value exceeds the allowable boundary of the stable window, it indicates that the fluctuation within the adjacent stable window is too large. Continuing to use it as a benchmark will reduce the reliability of the residuals. In this case, the adjacent stable window is marked as unusable. Subsequently, an adjacent stable window is reselected based on the source validity distribution, and the robust baseline residuals are updated based on the reselected adjacent stable window. The updated robust baseline residuals retain both the old and new window identifiers. When generating cross-source conflict in step S4, it can identify whether a window switch has occurred for the residual source.

[0066] Thus far, step S3 has completed the effective source screening, the construction of the nearest stable window, the generation of local benchmarks and scale quantities, and the update of robust baseline residuals, so that the reading deviation of the current coordinate interval has a comparable robust expression.

[0067] In step S4, the core task is to cross-validate the robust baseline residual with the execution side, the delivery side, and the coordinate attribution side, generate the cross-source conflict degree and the quality gating confidence, and output the gating state that the current control allows entry.

[0068] In this embodiment, cross-source conflict degree can be generated based on robust baseline residual and coordinate state chain. The cross-source conflict degree uses the correspondence between post-spray pattern reading anomalies and execution feedback as the judgment object, and reads the anomaly direction and duration from different sources through the coordinate state chain. Specifically, the coordinate state chain can identify the post-spray pattern reading anomaly amount, nozzle trigger confirmation deviation, response current deviation, speed fluctuation, encoder jump, and position residual window state within the current coordinate interval. The post-spray pattern reading anomaly amount is mainly formed by the robust baseline residual direction and the discontinuous nature of the post-spray pattern reading continuity; the nozzle trigger confirmation deviation is mainly formed by the deviation between the trigger command time and the confirmation return time; the response current deviation is mainly formed by the deviation between the channel response current stable segment and the historical confirmation segment; speed fluctuation and encoder jump originate from the transport state; and the position residual window state originates from the compensation coordinate in step S2.

[0069] In some embodiments of this application, the reading side, execution side, transport side, and coordinate attribution side each form independent conflict quantities. Each conflict quantity is jointly generated by the state source, deviation direction, and duration window within the current coordinate interval. Specifically, the post-jetting pattern reading anomaly quantity that is consistent with the direction of the robust baseline residual and satisfies the residual amplitude boundary can be classified into the reading side conflict quantity. The residual amplitude boundary can be jointly determined by the historical effective residual quantile value of the same plate type and the current adjacent stable window scale. Only reading anomalies with consistent direction and continuously exceeding the boundary are included in the reading side conflict quantity. The nozzle trigger confirmation deviation and response current deviation are classified into the execution side conflict quantity. The execution side conflict quantity requires that the trigger confirmation deviation and response current deviation have a time correspondence within the same channel or adjacent channels. The speed fluctuation and encoder jump are classified into the transport side conflict quantity. The transport side conflict quantity is used to indicate whether the reading deviation may be due to the instability of the plate movement. The position residual out-of-window state from different sources within the same current coordinate interval is classified into the coordinate attribution conflict quantity. The coordinate attribution conflict quantity is used to determine whether multiple sources are mismatched to the same current coordinate interval.

[0070] Based on this, a quaternary conflict quantity is generated according to the conflict quantity on the read side, the conflict quantity on the execution side, the conflict quantity on the transmission side, and the conflict quantity at the coordinate attribution. Then, a weighted sum is performed according to the source weight and coordinate attribution weight of each conflict quantity in the quaternary conflict quantity to obtain the cross-source conflict degree. The cross-source conflict degree can adopt a nonlinear compression form with source reliability and a read-execution coupling term; when both the read side and the execution side are enhanced simultaneously, the coupling term will improve the conflict response, while a single low-amplitude conflict still maintains a limited impact. For example, the first... Cross-source conflict degree corresponding to the current coordinate interval It can be written as: , in, , , and They represent the first The current coordinate interval corresponds to the read-side conflict, execution-side conflict, transmission-side conflict, and coordinate attribution conflict. , , and They represent the first The conflict participation coefficients of the reading side, execution side, transmission side, and coordinate attribution side within the current coordinate interval. This represents the co-directional coupling coefficient between read-side and execute-side conflict quantities. Indicates the first The conflict scale within the current coordinate interval is generated jointly by the source weight and the coordinate attribution weight. This represents the conflict scale stability constant. Cross-source conflict degree is incorporated into the quality gating confidence level. Single-source conflicts on the read side, related conflicts on the execution side, and coordinate attribution conflicts will trigger different gating processes.

[0071] Furthermore, the source weights are dynamically updated based on their actual usability within adjacent coordinate allocation windows. The source weights can be updated according to the number of times each source in the valid source set is available and isolated within adjacent coordinate allocation windows. The number of available times comes from the number of times the source enters the valid source set within the adjacent coordinate allocation window; the number of isolated times comes from the number of times the source enters the isolated source sequence or the source sequence awaiting review; and the position deviation from the mean is used to reflect the stability of the source relative to the plate travel reference. For example, the source weight update relationship can be written as: , in, Indicates the first The state source is in the first The updated source weight for the current coordinate range This indicates the number of times the state source is available within the adjacent coordinate attribution window. This indicates the number of times the source of this state is isolated within the adjacent coordinate attribution window. and Indicates the source weight smoothing parameter. This represents the mean of the location residuals of the source of this state within the adjacent coordinate assignment window. This indicates that the location corresponding to the source of this state deviates from the tolerance scale. This indicates a deviation from the stability constant. For example, the source weight update rate for the inlet plate state and the conveying state can be set to be slower, while the source weight for the post-spraying graphic reading state can be reduced more quickly when consecutive reading-side conflicts occur.

[0072] In some embodiments of this application, the coordinate attribution weight can be updated based on the persistence state of the position residual window within the current coordinate interval according to the coordinate attribution conflict amount. The persistence state of the position residual window is determined based on the number of windows through which the compensated coordinates continuously cross the allowable deviation of coordinate attribution; the more persistent windows, the lower the coordinate attribution weight. Further, a state deviation deduction is generated based on the deviation between the robust baseline residual corresponding to the current coordinate interval and the residual trigger boundary. This deduction reflects how far the current robust baseline residual is from the trigger boundary; the deduction is smaller when the residual has just crossed the boundary, and stronger when the residual continues to increase and cross-source conflicts increase synchronously.

[0073] Based on this, the quality gating confidence level is obtained by considering cross-source conflict degree, effective source set, current coordinate interval, and robust baseline residual, while preserving nonlinear suppression relationships. Specifically, different diversion methods are used for single-source conflicts on the read side and related conflicts on the execution side. When the conflict quantity on the read side meets the conflict condition, but the conflict quantity on the execution side, the conflict quantity on the transmission side, and the conflict quantity at the coordinate attribution side do not meet the conflict condition, the source weight corresponding to the conflict quantity on the read side can be reduced, and the source corresponding to the conflict quantity on the read side can be written into the verification waiting source sequence. The verification waiting source sequence records the original curve of the read window, the read continuity mark, and the illumination status, without changing the current control vector, only prompting subsequent verification acquisition. When the conflict quantity on the read side and the conflict quantity on the execution side meet the same-direction conflict condition, but the conflict quantity at the coordinate attribution side does not meet the conflict condition, the current coordinate interval is marked as the execution-side related conflict interval. The same-direction conflict condition requires that the direction of the robust baseline residual, the direction of nozzle trigger confirmation deviation, and the direction of response current deviation can correspond within the same channel or adjacent channel group. Finally, the quality gating confidence is obtained based on the cross-source conflict degree, valid source set, verification waiting source sequence, and state deviation deduction corresponding to the execution-side relevant conflict interval. Simultaneously, the execution-side relevant conflict interval allows for more cautious control vector solving, but the control vector change boundary is further restricted by step S5. For example, the first... Quality gate confidence level corresponding to each current coordinate interval It can be written as: , in, This indicates a deviation from the deduction amount. and They represent the first The residual trigger boundary and the corresponding effective source support for the current coordinate interval. and These represent the gating stability constant and the residual stability constant, respectively. The higher the cross-source conflict level, the greater the state deviation deduction, or the closer the robust baseline residual is to the anomalous amplification range, the lower the quality gating confidence level will be. The quality gating confidence level serves as the control entry decision quantity and, together with the current coordinate interval, determines the gating state.

[0074] Based on this, the gating state is determined according to the quality gating confidence level and the current coordinate interval. The current coordinate interval is divided into an entry transition zone, a stable control zone, a boundary avoidance zone, and a tail convergence zone, with different gating boundaries for each control zone. The entry transition zone requires a higher quality gating confidence level to allow control, the stable control zone allows for small, continuous corrections, the boundary avoidance zone applies stronger shielding constraints to the nozzle channels near the plate boundary, and the tail convergence zone prioritizes maintaining smooth control.

[0075] In some embodiments of this application, a coordinate state chain quality gating model can also be used in step S4. This model retains the above rule calculations and refines the source weights, state deviation deductions, and quality gating confidence based on historical review samples. The model input mainly consists of state nodes within the current coordinate interval. The node length can be 32 consecutive current coordinate intervals. A single node includes an inlet state integrity marker, a boundary positioning confidence marker, a delivery speed stability marker, a nozzle channel feedback integrity marker, a post-spray graphic reading continuity marker, a coordinate attribution consistency marker, a response delay residual, a robust baseline residual, a quaternion conflict quantity, and a current control area code. For example, the input feature dimension can be set to 64, where the coordinate state chain basic fields occupy 18 dimensions, the nozzle channel execution state occupies 16 dimensions, the delivery state occupies 8 dimensions, the post-spray graphic reading state occupies 14 dimensions, and the control area and review waiting state occupies 8 dimensions.

[0076] Specifically, the coordinate state chain quality gating model includes a state embedding layer, a time-delay residual memory layer, a cross-source conflict discrimination layer, and a gated readout layer. The state embedding layer employs a two-layer fully connected structure: the first layer maps the 64-dimensional input to a 96-dimensional hidden representation, and the second layer maps it to a 64-dimensional state representation, using a hyperbolic tangent activation function to limit outlier amplification. The time-delay residual memory layer uses a one-dimensional causal convolution with a kernel length of 5 and dilation rates of 1, 2, and 4 to cover common response delay spans between the inlet state, nozzle channel execution state, and post-spraying graphic reading state. The cross-source conflict discrimination layer inputs the read-side conflict quantity, execution-side conflict quantity, delivery-side conflict quantity, and coordinate attribution conflict quantity into independent gating branches. Each branch outputs a 16-dimensional conflict representation, which is then concatenated with the source weights and coordinate attribution weights to form an 80-dimensional comprehensive representation. The gated readout layer outputs two results: a corrected quality gating confidence score and a priority for reviewing the waiting source sequence. For example, the core encoding relationship of the model can be written as: , in, Indicates the first The model's hidden representation for the current coordinate interval. and They represent the first The state node input vector and the four-element conflict input vector for the current coordinate interval. and These represent the training weight matrices corresponding to the state node input vector and the four-element conflict input vector, respectively. This represents the bias vector of the model's hidden representation. The model's hidden representation then enters the gated readout layer and, together with the quality gate confidence obtained from rule computation, forms the final gated state.

[0077] Furthermore, training samples are drawn from batch process files, candidate event sequences, and verification waiting source sequences, but disabled sample indices corresponding to invalid verification confirmations and control oscillations are not included in the training set. Training samples are based on continuous coordinate state chain segments of the same board material. The training set, validation set, and test set are divided by batch, with proportions of 70%, 15%, and 15%, respectively. Sample labels are jointly determined by the verification confirmation results and the continuous changes in the graphic readings after control injection. Samples whose execution-side conflict intervals have recovered and stabilized are labeled as valid control samples; samples with conflict on the reading side and stable execution feedback are labeled as reading verification samples; and samples with continuous out-of-bounds coordinate attribution conflicts are labeled as coordinate frozen samples. When there are insufficient abnormal samples, only constrained perturbations are allowed on response delay, nozzle channel confirmation delay, local missing lines in post-injection graphic readings, conveyor speed perturbations, and board edge offsets. The perturbation amplitude must not exceed the allowable range of the equipment calibration and the order of the coordinate state chain must be maintained.

[0078] In some embodiments of this application, the model training employs the AdamW optimizer, with an initial learning rate of 0.001, a batch size of 64, and 120 training epochs. Training stops when the validation set gating misclassification rate fails to decrease for 12 consecutive epochs. The loss function simultaneously constrains the quality-gated confidence fitting, controls vector smoothing, and addresses cross-source conflict boundaries; the loss function can be written as: , in, This represents the training loss of the coordinate state chain quality-gated model. Indicates the number of training samples. Indicates the first Verification and confirmation confidence labels for each training sample. This indicates the gating confidence level of the model's prediction quality. This represents the control smoothing constraint coefficient. Indicates the first The control vector corresponding to each training sample This indicates the corresponding verification and confirmation reference control vector. Represents the stability constant of the control vector. Represents the cross-source conflict constraint coefficient. Indicates the first The cross-source conflict degree corresponding to each training sample This represents the corresponding verification and confirmation conflict boundary. The quality-gated fitting term of the training loss ensures that the model output fits the verification results, the control smoothness constraint limits the output from having an excessive impact on the control vector, and the cross-source conflict constraint ensures that single-source conflicts on the read side and related conflicts on the execution side remain separable within the model. During online inference, the model only updates the gated readout layer parameters with small steps on samples that have been verified and confirmed. The state embedding layer and the time-delay residual memory layer are frozen in the same batch to prevent unconfirmed samples from changing the online control boundary.

[0079] Thus far, step S4 has formed cross-source conflict representations for the reading side, execution side, transmission side, and coordinate attribution side, and obtained the quality gating confidence through rule gating and model refinement, providing boundary conditions for the output control vector in step S5.

[0080] In step S5, the core task is to convert the robust baseline residuals into control vectors constrained by the region boundary and hysteresis conditions, provided that the gating state allows, and to determine whether to include them in the batch process file based on the verification results.

[0081] In this embodiment, the control entry point simultaneously checks the quality gating confidence, cross-source conflict degree, and gating status. If the quality gating confidence meets the control conditions and the cross-source conflict degree does not exceed the conflict limit, it determines whether the gating status is an allowable control state. The control conditions are primarily determined by the current control area, the distribution of source validity, and the batch history review results. The conflict limit is primarily determined by the relevant conflict interval on the execution side and the duration of coordinate-attributed conflict. Only when all three conditions are met does the current coordinate interval enter the control vector solution. When the gating status is an allowable control state, a control vector can be generated based on the robust baseline residual, gating status, and quality gating confidence, and the control quantity is output according to the control vector. The control quantity ultimately acts on nozzle channel selection, injection trigger phase, pulse width, atomization intensity, channel shielding, delivery speed correction, and execution channel gain adjustment.

[0082] Specifically, solving for the control vector requires first identifying the control region where the coordinate interval lies. The current control region can be determined based on the current coordinate interval, including the inlet transition zone, stable control zone, boundary avoidance zone, and tail convergence zone. Then, the control vector change boundaries and hysteresis conditions are determined based on the current control region. The inlet transition zone restricts sudden nozzle channel switching, the stable control zone allows for small, continuous corrections, the boundary avoidance zone strengthens channel shielding and adjacent channel constraints, and the tail convergence zone limits trigger duration and delivery speed corrections.

[0083] Based on this, when the gated state is an allowable control state, the residual correction component is obtained according to the robust baseline residual and the quality gate confidence. The residual correction component comprehensively considers the robust baseline residual direction, the quality gate confidence, the current control region gain, and the cross-source conflict suppression amount; the higher the quality gate confidence and the lower the cross-source conflict, the closer the residual correction component is to the region's allowable gain; when the quality gate confidence just exceeds the control condition, the residual correction component is compressed. For example, the first... The residual correction component corresponding to the current coordinate interval It can be written as: , in, This indicates the feedback adjustment step size corresponding to the current control region. Indicates the first The control gain matrix for the current coordinate interval under the current control region. For example, the feedback adjustment step size of the stable control region can be 0.08-0.15, the entry transition region and the tail convergence region can be 0.03-0.08, and the boundary avoidance region is further reduced when approaching the edge of the plate.

[0084] Further, a control vector to be executed is generated based on the previous control vector and the residual correction component. This control vector is then subjected to amplitude limiting processing according to the control vector change boundary and hysteresis condition to obtain the final control vector. For example, the amplitude limiting processing can employ a region projection relationship, and the control vector output relationship can be written as: , in, Indicates the first The control vector corresponding to the current coordinate interval Indicates the first The region projection function of the current coordinate interval under the current control area includes the nozzle channel selectable boundary, injection trigger phase boundary, pulse width boundary, atomization intensity boundary, channel shielding boundary, delivery speed correction boundary and execution channel gain boundary; This represents the control vector corresponding to the previous current coordinate interval. and They represent the first The lower and upper limits of control for the current coordinate interval within the current control region. This represents the set of hysteresis boundaries. Simultaneously, control quantities are output according to the control vector, including nozzle channel selection, injection trigger phase, pulse width, atomization intensity, channel shielding, delivery speed correction, and execution channel gain adjustment. Among them, nozzle channel selection and channel shielding are jointly determined based on the current control region and feedback from adjacent channels. Injection trigger phase and pulse width are updated in small steps based on residual correction components. Atomization intensity and execution channel gain adjustment are subject to additional constraints from the relevant conflict intervals on the execution side. Delivery speed correction is only output when the conflict amount on the delivery side does not exceed the control boundary, and if any control quantity exceeds the boundary, it is recovered according to the nearest executable boundary.

[0085] Furthermore, the control vector and the current coordinate interval are bound together to generate the control result. The control result fields include the current coordinate interval, control vector, gating state, quality gating confidence, cross-source conflict degree, current control region, region projection boundary, and output time.

[0086] It should be noted that after the control results are output, they need to enter the verification chain to write the verified and valid control results into the batch process file. Specifically, a verification object can be generated based on the control vector and the current coordinate range in the control results. The verification object includes the continuous values ​​of the post-injection graphic reading before and after control, nozzle channel feedback, conveyor speed changes, and coordinate attribution status. Then, the verification order is determined according to the degree of cross-source conflict, including post-injection graphic reading verification, nozzle channel feedback verification, conveyor speed verification, and coordinate attribution verification. Among them, when the reading-side conflict contribution is the highest, the original curve of the post-injection graphic reading is verified first; when the execution-side conflict contribution is the highest, the nozzle channel feedback is verified first; when the conveyor-side conflict contribution is the highest, the conveyor speed is verified first; and when the coordinate attribution conflict contribution is the highest, the response delay confidence interval and compensation coordinates are verified first.

[0087] Based on this, review confirmation results are generated according to the review object and review order, including valid review, unidentified review source, and invalid review. Valid review requires that after the control result is output, the continuous reading of the post-spray graphic returns to the allowable range of the adjacent stable window within several subsequent current coordinate intervals, and that the nozzle channel execution state and delivery state do not generate new cross-source conflicts. Unidentified review source indicates that there is a corresponding relationship between the control result and the reading improvement, but the source evidence is insufficient. Invalid review indicates that the control result has not improved the continuous reading of the post-spray graphic, or has triggered new execution-side related conflict intervals.

[0088] Furthermore, when the verification confirmation result is valid, the control vector, current coordinate interval, quality gate confidence level, and verification confirmation result are written into the batch process file. The batch process file is indexed by plate type, plate width, speed range, nozzle channel configuration, and current control area. The fields written include the control vector, current coordinate interval, quality gate confidence level, verification confirmation result, verification time, and source weight status. When the verification confirmation result indicates that the verification source is unknown, the control result is written into the candidate event sequence. The candidate event sequence is only used for subsequent anomaly tracing and does not participate in the next batch's initial control value update. When the verification confirmation result indicates that the verification is invalid or the control result causes control oscillation, the current coordinate interval is written into the disabled sample index. The data corresponding to the disabled sample index does not participate in the next batch's initial control value update. The disabled sample index also includes the plate number, control vector direction, oscillation boundary, and verification confirmation result. Subsequent online updates will skip the state chain segment corresponding to this index.

[0089] In this process, control oscillation includes the state where the control vector directions of adjacent batches corresponding to the same current coordinate interval alternate and the amplitude of the control vector exceeds the oscillation boundary. This can be determined by the consistency of the control vector direction and the change in amplitude of adjacent batches within the same current coordinate interval. For example, the oscillation constraint condition can be expressed as: ,in, Indicates the first The first batch The control vector corresponding to the current coordinate interval This represents the control vector corresponding to the same current coordinate interval in the previous batch. This represents the oscillation boundary corresponding to the current control region. The smaller this ratio, the closer the control vector directions are to being opposite in adjacent batches; when the control vector amplitudes simultaneously cross the oscillation boundary, the current coordinate interval enters the disabled sample index.

[0090] Furthermore, the batch process file only accepts control results that have been verified and confirmed to be valid. The initial control values ​​for the next batch can be updated in small steps based on the verification results and the quality gate confidence level. The update relationship can be written as: , in, This indicates the process file parameters for the next batch. This indicates the process file parameters for the current batch. Indicates the file update step size. Indicates the first The current coordinate interval corresponds to a verification and confirmation flag; the verification and confirmation flag comes from the verification and confirmation result, the quality gating confidence comes from step S4, and the control vector comes from the output of the current step. When the verification and confirmation flag is invalid or of unknown origin, the file parameters remain unchanged; when the verification and confirmation flag is valid and the quality gating confidence is high, the control vector is adjusted in small steps to correct the initial control value of the next batch.

[0091] Thus, step S5 completes the closed-loop process from gating status confirmation, residual correction, control vector limiting to write-back of verification results, so that the control output is jointly limited by quality constraints, regional constraints and batch verification boundaries.

[0092] In summary, this method incorporates the inlet plate state, nozzle channel execution state, conveying state, and post-spraying graphic reading state into the same plate travel coordinate system for time delay compensation and state chain construction. It combines robust baseline residuals within the adjacent stable window, cross-source conflict degree, and quality gating confidence to form control output constraints, thereby achieving multi-source state alignment, abnormal source diversion, control vector amplitude limiting output, and verification result write-back. This method can improve the consistency of control basis, the stability of control quantity adjustment, and the traceability and controllability of continuous batch glue application results under conditions of fluctuating conveying speed, nozzle response lag, and post-spraying graphic reading delay.

[0093] It should be noted that, although the embodiments in this application are based on... Figure 1 The steps are described sequentially, but this does not mean that the steps must be performed in a strict order. The reason this embodiment follows this order is... Figure 1The order in which each step is described is intended to facilitate understanding of the technical solutions of the embodiments of this application by those skilled in the art. In other words, the step numbers are only used to distinguish different steps and do not constitute a limitation on the execution order of the steps; the specific execution order of each step can be appropriately adjusted according to actual needs, functional requirements, and the inherent logic in actual application scenarios.

[0094] In some embodiments of this application, a method for controlling the application of adhesive to a cleanroom panel is applied to the adhesive application station of a continuous production line for cleanroom panels. A code reading unit and an edge positioning unit are set before the entrance baseline, an encoder is configured on the conveying mechanism, and the spraying area is set... Each nozzle channel has a line scan reading unit located behind the spray area. The production line controller uses a length of... Width is The purification plate is the object of treatment, and the sampling interval is set to The coordinate interval length of the stable control region is set to A complete implementation process involves continuous steps including board material entry, status alignment, residual judgment, quality gate control, control output, and batch write-back.

[0095] A complete implementation process may include the following steps: Step 1: Collect the states of the ingress side, execution side, transmission side, and read side at the same sampling time, and form a multi-source state packet. The source set corresponding to each sampling time is ,in Indicates the condition of the inlet material. Indicates the nozzle channel execution status. Indicates the conveying status. This indicates the status of the image read after spraying; the production line controller writes the source number, original acquisition timestamp or feedback arrival time, original projection position, original status field, and initial source quality screening value into the multi-source status packet: In this embodiment, the nozzle channel execution state source The feedback arrival time is Corresponding to the original projection position The original projection position corresponding to the previous sampling time. Permissible scale of change between adjacent locations Allowable scale of change between adjacent original times Position stability constant The constant of time Substituting the source quality initial screening relationship: Since 0.64 is higher than the on-site set lower limit of 0.50, the source... Proceed to subsequent coordinate compensation; if the same source is below this lower limit, the source will only be retained as the original record to be verified and will not participate in the control judgment of the current coordinate interval.

[0096] Step 2: Complete the conversion of the plate travel reference and the calculation of the response delay residual according to the conveying status.

[0097] The entry baseline trigger time is Initial coordinates of the entrance baseline Encoder count at the 186th sampling time Encoder count when entry is triggered encoder unit distance resolution Average conveying speed .

[0098] The production line controller uses a time coordinate mapping relationship to calculate the source. The sheet travel reference at this feedback arrival time: Based on this, at the velocity stability constant At that time, source The response delay residual is: The response delay residual sequence obtained from the same source within a continuous stable window is: The bit value is The median absolute deviation is ,Pick Therefore, the updated response delay confidence interval in this embodiment is: The online response delay is estimated using the center of this interval. Then the compensated coordinates of the plate movement are .

[0099] At the same sampling time, the inlet compensation coordinates, nozzle trigger compensation coordinates, and graphic reading compensation coordinates all fall within The corresponding coordinates belong to the window, and the center travel coordinates of that window. Allowable deviation The coordinate attribution window is written as The production line controller then generates a coordinate status chain node that includes inlet status integrity markers, conveyor speed stability markers, nozzle channel feedback integrity markers, and post-spray graphic reading continuity markers.

[0100] Step 3: Select a nearby stable window on the coordinate state chain and generate robust baseline residuals. Current coordinate interval Located in the stable control region, candidate neighbor windows are selected from both the front and back sides. The coordinate intervals are selected, and the entry transition zone, tail convergence zone, boundary avoidance zone, and intervals where the source validity does not meet the conditions are removed. The continuous amount of post-jet graphic reading output by the line scan reading unit within the adjacent stable window is [value missing]. The production line controller forms a benchmark sample set. The local benchmark value of this set. The median absolute deviation is The interquartile range is converted to... Scale stability constant Therefore, the scale quantity is: , Continuous reading of the current coordinate range The robust baseline residual is This value indicates that the current continuous reading of the refreshed image is lower than the local benchmark of the adjacent stable window, and the deviation exceeds the residual trigger boundary. However, this deviation still needs to be verified together with the status of the execution side, the delivery side, and the coordinate attribution side.

[0101] Step four: Calculate the cross-source conflict degree and quality gating confidence degree based on the quaternary conflict quantities. The production line controller reads the post-spraying graphic reading anomalies, nozzle trigger confirmation deviations, response current deviations, speed fluctuations, and position residual out-of-window status within the current coordinate range. In this embodiment, the side conflict quantity is read. Execution side conflict volume Conveyor-side collision volume Coordinate attribution conflict quantity The corresponding source participation coefficients are respectively , , and Read-execute coupling coefficient Conflict scale Conflict stability constant Substituting the values ​​into the cross-source conflict calculation, we obtain... Furthermore, the effective source of support Deduction amount for status deviation Gated stability constant residual stability constant The quality gate confidence level is: , because It is higher than the lower limit of the stable control region gate by 0.65, and If the value is below the conflict limit of 0.60, the current gating state is determined to be an allowed control state; if the value is below the limit in the next two windows... Increase or If the value decreases, the same coordinate interval will enter the source sequence for review.

[0102] Step 5: Construct and invoke the coordinate state chain quality gating model. The model backbone used in this embodiment includes a state chain feature projection layer, a quaternary conflict projection layer, a gating hidden layer, a gating confidence output head, and a control reference output head. State chain features 24-dimensional input is used, including inlet integrity, boundary confidence, conveyor speed stability marker, nozzle feedback integrity marker, read continuity marker, and coordinate attribution consistency marker. , , The current control region code and the number of source isolations in the three most recent coordinate intervals; quaternary conflict characteristics. Using 4D input, the order is as follows , , and State chain feature projection matrix The size is 16×24, a four-element conflict projection matrix. The size is 16×4, and the bias vector The length is 16, and the gated hidden layer is generated according to the following formula: The gated confidence output head is... Output prediction gate value Control the reference output head output ,in This corresponds to nozzle channel selection, injection trigger phase, pulse width, atomization intensity, channel shielding, delivery speed correction, and execution channel gain adjustment. The model training objectives focus on gating value fitting, control reference offset constraints, and cross-source conflict boundary constraints; the original graphical readout curves in the state chain are only used as feature inputs and verification references, not as direct output control quantities. To adapt to the cleanroom plate adhesive application control task, the model incorporates the inlet transition zone, stable control zone, boundary avoidance zone, and tail-end convergence zone as region codes. This allows the same residual magnitude to correspond to different lower thresholds and control boundaries in different regions; simultaneously, the model incorporates statistics from the verification waiting source sequence and the isolated source sequence into the model. This causes low-quality sources to receive a lower weight in the next inference. During online inference, the model only updates... , And control the reference output without rewriting the coordinate state chain that has already been formed. , , and Intermediate quantities, etc.

[0103] It should be noted that, in this embodiment, the coordinate state chain quality gating model retains the state chain feature projection layer and the quaternary conflict projection layer, compressing the general 64-dimensional input into 24-dimensional state chain features and 4-dimensional quaternary conflict features; this embodiment is used to illustrate the gating readout process and does not limit the aforementioned 64-dimensional general input structure.

[0104] Step Six: Train the coordinate state chain quality gating model using historical batch data. Training samples are drawn from the cleanroom panels with complete verification results from the most recent 80 batches, extracted stratified according to panel type, speed range, and nozzle channel configuration. The training, validation, and test sets are divided by batch to avoid adjacent panels from the same batch appearing in different sets simultaneously. A single training sample consists of a current coordinate range and its adjacent stable window, including multi-source state packets, coordinate state chain nodes, a baseline sample set, quaternion conflict variables, control vectors, and verification confirmation results. Data augmentation applies limited perturbations only to response delay, partial missing lines in line scan, minor perturbations in conveying speed, and edge positioning offset. For example, the response delay perturbation is controlled within... Within the range, the percentage of missing lines in line scanning is controlled to within 8%, and the edge positioning offset is controlled to within... Within this framework, all enhanced samples must maintain continuous board movement coordinates. The training loss consists of three parts: gating value fitting, control reference offset constraint, and cross-source conflict boundary constraint. In this embodiment, the batch size is 64, the initial learning rate is 0.001, and the maximum number of training epochs is 120. Training stops when the validation set gating misfire rate does not decrease for 10 consecutive epochs. After going live, only samples that have been verified and confirmed to be valid are allowed to update the gating confidence output head and control reference output head with small steps. The state chain feature projection layer and the quaternary conflict projection layer remain frozen within the current batch. Samples with unclear verification sources or invalid verification are entered into the candidate event sequence or disabled sample index and do not participate in the control initial value update of the next batch.

[0105] Step 7: Generate control vectors and output control quantities under permitted control conditions. The current coordinate interval is in the stable control region, and the previous control vector... The corresponding injection trigger phase is Pulse width multiplier is 1.00, atomization intensity multiplier is 1.00, and delivery speed is corrected to... Take feedback and adjust the step size. Local control gain Calculated based on residual correction components The control vector to be executed passes through the lower control limit of the stable control region. Control upper limit and hysteresis boundary After projection, the control vector is obtained. This control vector adjusts the pulse width multiplier of the corresponding nozzle channel from 1.00 to 1.06, delaying the injection trigger phase. Atomization intensity ratio remains at 1.00, and delivery speed correction remains unchanged. The channel near the edge of the sheet metal remains shielded. The production line controller modifies this control vector with the current coordinate range. The control results are bound together, and the nozzle channel feedback and post-spray graphic reading status are read at the subsequent two sampling times.

[0106] Step 8: Update the batch process file based on the verification results. After control execution, the nozzle channel trigger confirmation returned to stable operation, and no abnormal jumps occurred in the response current. The subsequent continuous readings in the coordinate intervals recovered to 0.938 and 0.942, both falling within the allowable range of the adjacent stable window. The verification result is recorded as follows. Current batch process file parameters Archive update step size Substituting the file update relationship, we get: , Production line controller will , , , The current coordinate range, verification confirmation result, and status chain index are written into the batch process file. When the next batch with the same plate type, speed range, and nozzle channel configuration starts, the production line controller calls... This serves as the initial control value for the coordinate range. If the verification result is that the source of verification is unknown, the control result is written into the candidate event sequence. If the verification is invalid or adjacent batches show alternating directions and amplitudes exceeding the oscillation boundary, the corresponding state chain segment is written into the disabled sample index.

[0107] Through the above complete implementation process, the inlet status, nozzle execution status, conveying status, and post-spraying graphic reading status are unified under the board travel coordinate system. Combined with robust baseline residuals, cross-source conflict degree, quality gate confidence degree, and review write-back boundary, a traceable closed-loop control process is formed. This can reduce source mismatch, reading noise amplification, and control vector jumps when conveying speed fluctuations, nozzle response lag, and line scan reading disturbances coexist. It also improves the coordinated stability between nozzle channel selection, trigger phase, pulse width, and conveying correction, and allows the control experience that has been verified to be effective to be included in the batch process file. This improves the control consistency, anomaly attribution reliability, and batch-to-batch traceability and controllability of the continuous clean board gluing process.

[0108] It should be noted that the step numbers in the above embodiments, such as step one, step two, etc., are only used to distinguish different processing stages and do not limit the steps to be executed in the order of these numbers. The specific execution order of each step can be adjusted according to its functional requirements and the inherent logic in the actual application scenario. The above step numbers should not be interpreted as a limitation on the implementation process of the embodiments of this application.

[0109] like Figure 2 As shown, the following is an embodiment of a cleanroom panel gluing control system provided in this application. This cleanroom panel gluing control system and the cleanroom panel gluing control methods in the above embodiments belong to the same inventive concept. For details not described in detail in the embodiments of the cleanroom panel gluing control system, please refer to the embodiments of the cleanroom panel gluing control methods described above.

[0110] Based on the same concept, another embodiment of this application provides a cleanroom panel adhesive application control system, comprising: Status acquisition unit 1 is used to acquire the status of the inlet plate, as well as the nozzle channel execution status, conveying status and post-spraying graphic reading status corresponding to the status of the inlet plate. Coordinate construction unit 2 is used to perform time delay compensation on the inlet plate state, nozzle channel execution state and post-spraying graphic reading state according to the conveying state, obtain the plate travel coordinates, and construct a coordinate state chain according to the plate travel coordinates. The residual generation unit 3 is used to read the effective source set and effective state quantities from the coordinate state chain, and generate robust baseline residuals based on the effective state quantities within the adjacent stable window. The gating determination unit 4 is used to generate cross-source conflict degree based on robust baseline residuals and coordinate state chain, obtain quality gating confidence based on cross-source conflict degree, effective source set, current coordinate interval and robust baseline residuals, and determine gating state based on quality gating confidence and current coordinate interval; The control output unit 5 is used to generate a control vector based on the robust baseline residual, the gate state, and the quality gate confidence when the quality gate confidence meets the control conditions, the cross-source conflict degree does not exceed the conflict limit, and the gate state is an allowable control state. It then outputs the control quantity according to the control vector and writes the verified and confirmed effective control results into the batch process file.

[0111] The above-disclosed embodiments are merely preferred embodiments of this application, but this application is not limited thereto. Any changes, improvements, and modifications made by those skilled in the art without departing from the principles of this application, without inventive effort, shall fall within the protection scope of this application.

Claims

1. A method for controlling the application of adhesive to a cleanroom panel, characterized in that, include: Step S1: Obtain the inlet plate status, and the nozzle channel execution status, conveying status, and post-spraying graphic reading status corresponding to the inlet plate status; Step S2: Perform time delay compensation on the inlet plate state, nozzle channel execution state, and post-spray graphic reading state according to the conveying state to obtain the plate's travel coordinates, and construct a coordinate state chain based on the plate's travel coordinates; wherein, the step of performing time delay compensation on the inlet plate state, nozzle channel execution state, and post-spray graphic reading state according to the conveying state to obtain the plate's travel coordinates specifically includes: generating a plate travel reference according to the conveying state, the plate travel reference including the time coordinate mapping relationship determined according to the conveying speed and encoder state; obtaining the inlet projection position, the inlet projection position, and the post-spray graphic reading state respectively according to the inlet plate state, nozzle channel execution state, and post-spray graphic reading state. The nozzle trigger projection position and the graphic reading projection position are determined. Based on the coordinate deviations of the inlet projection position, nozzle trigger projection position, and graphic reading projection position relative to the plate travel reference, the response delay residual is obtained. When the response delay residual satisfies the same-direction offset condition within a continuous stable window, the response delay confidence interval is updated based on the response delay residual. Delay compensation is performed on the inlet plate state, nozzle channel execution state, and post-spray graphic reading state according to the response delay confidence interval to obtain the plate travel coordinates. When the response delay confidence interval exceeds the allowable boundary, the corresponding source is written into the source sequence to be reviewed, and the corresponding source does not participate in the control judgment of the current coordinate interval. Step S3: Read the effective source set and effective state variables from the coordinate state chain, and generate robust baseline residuals within the nearest stable window based on the effective state variables; Step S4: Generate cross-source conflict degree based on robust baseline residuals and coordinate state chain; obtain quality gating confidence based on cross-source conflict degree, effective source set, current coordinate interval and robust baseline residuals; and determine gating state based on quality gating confidence and current coordinate interval. Step S5: When the quality gating confidence meets the control conditions, the cross-source conflict degree does not exceed the conflict limit, and the gating state is an allowable control state, generate a control vector based on the robust baseline residual, the gating state, and the quality gating confidence, and output the control quantity according to the control vector. Write the verified and confirmed effective control results into the batch process file.

2. The method for controlling the application of adhesive to a cleanroom panel as described in claim 1, characterized in that, Step S2, which involves constructing a coordinate state chain based on the plate's travel coordinates, specifically includes: The inlet compensation coordinates, nozzle trigger compensation coordinates, and graphic reading compensation coordinates are obtained based on the plate travel coordinates and the response delay confidence interval, respectively. A coordinate assignment window is generated based on the coordinate overlap relationship between the inlet compensation coordinates, the nozzle trigger compensation coordinates, and the graphic reading compensation coordinates. The following markers are determined according to the coordinate attribution window: inlet status integrity marker, boundary positioning reliability marker, conveying speed stability marker, nozzle channel feedback integrity marker, post-spray graphic reading continuity marker, and coordinate attribution consistency marker. When there are states within the same coordinate attribution window that do not meet the source validity condition, write the states that do not meet the source validity condition into the isolated source sequence; A coordinate state chain is generated based on the plate's travel coordinates, the coordinate ownership window, and the states within the coordinate ownership window that have not been written with an isolated source sequence.

3. The method for controlling the application of adhesive to a cleanroom panel as described in claim 2, characterized in that, Step S3, which involves generating robust baseline residuals within a nearby stable window based on effective state variables, specifically includes: Determine the source validity distribution of the current coordinate interval based on the valid state quantities; Candidate neighbor windows are determined from the coordinate state chain based on the source validity distribution, and the inlet transition region, tail convergence region, boundary avoidance region and coordinate intervals where the source validity does not meet the conditions are removed from the candidate neighbor windows to obtain the neighbor stable window; Read the continuous data of the post-spray graphic within the adjacent stable window to generate a set of benchmark samples; Local baseline values ​​and scale quantities are obtained from the baseline sample set, and robust baseline residuals are generated by reading continuous quantities, local baseline values, and scale quantities from the post-spraying graphic corresponding to the current coordinate interval.

4. The method for controlling the application of adhesive to a cleanroom panel as described in claim 3, characterized in that, The steps include obtaining local baseline values ​​and scale quantities from the baseline sample set, and generating robust baseline residuals by reading continuous quantities, local baseline values, and scale quantities from the post-jetting pattern corresponding to the current coordinate interval. The local benchmark value is obtained based on the median of the benchmark sample set; The scale quantity is obtained by reading the median absolute deviation or interquartile range of the continuous quantity relative to the local reference value from the sprayed graphic in the reference sample set. Based on the difference between the continuous quantity and the local reference value read from the post-spraying graphic corresponding to the current coordinate interval, the scale quantity is normalized to obtain the robust baseline residual. When the scale quantity exceeds the allowable boundary of the stability window, the neighboring stability windows are marked as unusable windows, and a new neighboring stability window is selected based on the source validity distribution. The robust baseline residuals are updated based on the newly selected nearest stable window.

5. The method for controlling the application of adhesive to a cleanroom panel as described in claim 2, characterized in that, Step S4, which generates cross-source conflict degree based on robust baseline residuals and coordinate state chains, specifically includes: Based on the coordinate state chain, identify the abnormal amount of post-spray graphic reading, nozzle trigger confirmation deviation, response current deviation, speed fluctuation, encoder jump and position residual out-of-window status within the current coordinate range; The abnormal reading amount of the post-jetting pattern that is consistent with the direction of the robust baseline residual and satisfies the residual magnitude boundary is classified as the reading-side conflict amount. The nozzle trigger confirmation deviation and response current deviation are classified into the execution-side conflict quantity; Speed ​​fluctuations and encoder jumps are classified as conveyor-side conflict quantities. The position residual window states from different sources within the same current coordinate interval are classified into coordinate attribution conflict quantities. A quaternary conflict quantity is generated based on the conflict quantity on the read side, the conflict quantity on the execution side, the conflict quantity on the transmission side, and the conflict quantity on the coordinate attribution side. The cross-source conflict degree is obtained by weighting and summing the conflict quantities according to the source weight and coordinate attribution weight of each conflict quantity in the quaternary conflict quantity.

6. The method for controlling the application of adhesive to a cleanroom panel as described in claim 5, characterized in that, Step S4, which involves obtaining the quality-gated confidence level based on cross-source conflict degree, valid source set, current coordinate interval, and robust baseline residuals, specifically includes: Update the source weights based on the number of times each source is available and isolated within the adjacent coordinate allocation window in the valid source set; Update the coordinate assignment weights based on the position residual window persistence state of the coordinate assignment conflict quantity within the current coordinate interval. A state deviation deduction is generated based on the deviation between the robust baseline residual and the residual trigger boundary corresponding to the current coordinate interval. When the conflict quantity on the read side meets the conflict condition, but the conflict quantity on the execution side, the conflict quantity on the transmission side, and the conflict quantity on the coordinate attribution do not meet the conflict condition, the source weight corresponding to the conflict quantity on the read side is reduced, and the source corresponding to the conflict quantity on the read side is written into the review waiting source sequence. When the conflict quantity on the read side and the conflict quantity on the execution side meet the same-direction conflict condition and the conflict quantity on the coordinate attribution does not meet the conflict condition, the current coordinate interval is marked as the execution-side related conflict interval, and the quality gating confidence is obtained based on the cross-source conflict degree, the set of valid sources, the sequence of sources waiting for review, and the state deviation deduction amount corresponding to the execution-side related conflict interval.

7. The method for controlling adhesive application on a cleanroom panel as described in claim 1, characterized in that, Step S5, which involves generating a control vector based on the robust baseline residuals, gating state, and quality gating confidence, and then outputting the control quantity according to the control vector, specifically includes: The current control region is determined based on the current coordinate interval, and the control vector change boundary and hysteresis condition are determined based on the current control region. The current control region includes the inlet transition region, the stable control region, the boundary avoidance region, and the tail convergence region. When the gated state is the allowed control state, the residual correction component is obtained based on the robust baseline residual and the quality gate confidence, and the control vector to be executed is generated based on the previous control vector and the residual correction component. The control vector to be executed is limited according to the control vector change boundary and hysteresis condition to obtain the control vector. The control quantity is then output according to the control vector. The control vector and the current coordinate interval are bound to generate the control result. The control quantity includes nozzle channel selection, injection trigger phase, pulse width, atomization intensity, channel shielding, delivery speed correction and execution channel gain adjustment.

8. The method for controlling the application of adhesive to a cleanroom panel as described in claim 7, characterized in that, Step S5, which involves writing the verified and confirmed control results into the batch process file, specifically includes: The verification object is generated based on the control vector and the current coordinate range in the control results, and the verification order is determined according to the cross-source conflict degree. The verification order includes post-spraying graphic reading verification, nozzle channel feedback verification, conveying speed verification, and coordinate attribution verification. The review confirmation results are generated based on the review object and review order. The review confirmation results include review confirmation valid, review source unknown, and review confirmation invalid. When the verification result is valid, the control vector, current coordinate range, quality gate confidence level and verification result will be written into the batch process file. When the review confirmation result is that the source of the review is unknown, the control result will be written into the candidate event sequence; When the verification result is invalid or the control result causes control oscillation, the current coordinate interval is written to the disabled sample index. The data corresponding to the disabled sample index will not participate in the update of the initial control value of the next batch. Control oscillation includes the state in which the control vector direction of the same current coordinate interval in adjacent batches alternates and the amplitude of the control vector exceeds the oscillation boundary.

9. A cleanroom panel adhesive application control system, used to implement the cleanroom panel adhesive application control method according to claim 1, characterized in that, The system includes: The status acquisition unit is used to acquire the status of the inlet plate, as well as the nozzle channel execution status, conveying status, and post-spraying graphic reading status corresponding to the status of the inlet plate. The coordinate construction unit is used to perform time delay compensation on the inlet plate state, nozzle channel execution state and post-spraying graphic reading state according to the conveying state, to obtain the plate travel coordinates, and to construct a coordinate state chain according to the plate travel coordinates. The residual generation unit is used to read the effective source set and effective state quantities from the coordinate state chain, and generate robust baseline residuals based on the effective state quantities within the adjacent stability window. The gating determination unit is used to generate cross-source conflict degree based on robust baseline residuals and coordinate state chain, obtain quality gating confidence based on cross-source conflict degree, effective source set, current coordinate interval and robust baseline residuals, and determine gating state based on quality gating confidence and current coordinate interval; The control output unit is used to generate a control vector based on the robust baseline residual, the gate state, and the quality gate confidence when the quality gate confidence meets the control conditions, the cross-source conflict degree does not exceed the conflict limit, and the gate state is an allowable control state. It then outputs the control quantity according to the control vector and writes the verified and confirmed effective control results into the batch process file.

Images