A nozzle self-cleaning process control method and system

Abstract

The application relates to the field of precision coating technology and discloses a nozzle self-cleaning process control method and system, which comprises the following steps: acquiring pressure time sequence data of the nozzle in a cleaning process; extracting trend features and stability features of the pressure time sequence data; identifying a cleaning stage of the current nozzle based on the trend features and the stability features; and executing a cleaning end point control action when it is determined that the nozzle is in a cleaning limit stage. The application effectively ensures the thoroughness of nozzle cleaning, thereby improving the stability and production efficiency of the precision coating process.

Description

Technical Field

[0001] This application relates to the field of precision coating technology, and more specifically, to a method and system for controlling the self-cleaning process of a nozzle. Background Technology

[0002] In precision coating processes, to maintain nozzle flow rate and adhesive rejection performance, automatic hot solvent rinsing and purging of the coating nozzles are typically performed between batch production runs. During rinsing, the system monitors the back pressure at the nozzle outlet in real time. Once the back pressure drops to a preset low-pressure threshold for cleaning completion and stabilizes for a period, the blockage is considered cleared. Solvent injection is then stopped, and pneumatic purging ends the cleaning cycle. In actual electronic manufacturing dispensing processes, due to the heat-sensitive properties of UV adhesives, nozzles are prone to producing residues of varying forms and bond strengths during heating and prolonged operation cycles. When nozzles continuously process adhesives with different aging levels, the blockage sometimes consists mainly of incompletely cured viscous clumps, sometimes of semi-charred hardened skin formed by overheating, and sometimes even a complex blockage state where viscous clumps and stubborn gels coexist. When the blockage is loose particles or simply viscous clumps, the shear force provided by solvent rinsing can quickly peel them off, causing the back pressure to drop sharply below the threshold, and the cleaning task is successfully completed.

[0003] However, when faced with complex blockages caused by a mixture of viscous micelles and stubborn gel, solvent flushing initially dissolves the outer viscous micelles, leading to a rapid and significant drop in back pressure. But when confronted with the slowly softening inner layer of stubborn gel, the back pressure drop gradient abruptly decreases and slowly approaches a limit slightly above a preset threshold. At this point, the control logic recognizes that the back pressure has not fallen below the threshold and cannot stop cleaning; however, because the back pressure drop is already extremely slow, it cannot reach the threshold within the preset maximum safe flushing time. Existing control logic based on a fixed back pressure threshold cannot distinguish between two states: essentially cleaned but with stubborn residue, and still effectively cleaning. This leads to the cleaning task making an incorrect choice between timeout failure and forced exit with residue.

[0004] To address the aforementioned issues, existing technologies urgently need improvement. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this application discloses a nozzle self-cleaning process control method and system, which aims to solve the technical problem in existing technologies where the inability to distinguish between the effective cleaning limit and the state of residual stubborn gel leads to the incorrect choice between timeout failure and forced exit with residue in the cleaning task.

[0006] The technical solution of this application is as follows: In a first aspect, this application discloses a nozzle self-cleaning process control method, comprising: Acquire pressure timing data of the nozzle during the cleaning process; Extract trend and stability features from pressure time-series data; Based on trend and stability characteristics, identify the current cleaning stage of the nozzle; When the nozzle is determined to be in the cleaning limit stage, the cleaning endpoint control action is executed.

[0007] This technical solution can effectively distinguish between cleaning limit states and continuous cleaning states by dynamically identifying cleaning stages rather than relying on a single fixed threshold, thereby avoiding cleaning failures caused by timeouts or misjudgments.

[0008] Optionally, the step of acquiring pressure timing data of the nozzle during the cleaning process includes: Extract the original discrete-time sequence of pressure from the nozzle at a preset sampling frequency; The original discrete-time pressure series is validated to remove outliers, resulting in a validated pressure series. The pressure sequence after verification is subjected to sliding window averaging filtering and real-time zero-point compensation to obtain pressure time series data.

[0009] Optionally, trend features include the decrease per unit time and the rate of decrease decay, and stability features include pressure slip variance; The steps for extracting trend and stability features from stress time series data include: Linear regression was performed on the pressure time series data within a preset short time window, and the slope of the fitted line was extracted as the decrease per unit time. Calculate the difference or ratio between the unit time decrease of the current short time window and the unit time decrease of the adjacent previous short time window to obtain the rate of decrease decay. The variance of pressure time series data within a preset long-term window is calculated to obtain the pressure sliding variance.

[0010] Optionally, the cleaning phases include a rapid removal phase, a continuously effective removal phase, an abnormal stagnation phase, and a cleaning limit phase; Based on trend and stability characteristics, the steps to identify the current nozzle cleaning stage include: When the decrease per unit time is between the preset first gradient threshold and the preset second gradient threshold, and the pressure sliding variance is less than the preset first stable threshold, the nozzle is determined to be in the continuous effective cleaning stage. When the absolute pressure value corresponding to the pressure time series data exceeds the preset acceptable residual tolerance zone and the decrease per unit time approaches zero; or when the pressure sliding variance exceeds the preset second stability threshold, the nozzle is determined to be in an abnormal stagnation stage. When the unit time drop is less than the preset first threshold, the drop rate decay is within the preset fluctuation range, and the absolute pressure value corresponding to the pressure time series data falls into the preset acceptable residual tolerance zone, the preset limit confirmation timer is started; during the operation of the limit confirmation timer, if the unit time drop and pressure sliding variance are continuously lower than the second threshold within the preset time window, the nozzle is determined to be in the cleaning limit stage. Otherwise, the nozzle is determined to be in the rapid cleaning phase.

[0011] Optionally, when it is determined that the nozzle is in an abnormal stagnation phase, the method further includes: If the absolute pressure value exceeds the acceptable residual tolerance zone and the pressure drop per unit time is zero, the liquid supply pump will stop and a fault signal will be output. If the pressure sliding variance fluctuates and the total flushing time reaches the preset safe duration threshold, an alternating gas-liquid impact action is executed, and the cleaning process is determined based on the pressure sliding variance after the impact to either continue cleaning or issue a cleanliness doubt warning.

[0012] Optionally, during the execution of the limit confirmation timer, the method further includes: If the absolute value of the decrease per unit time is detected to be greater than the first threshold, the limit confirmation timer is suspended and the disturbance identification state is entered. In the disturbance identification state, a preset disturbance observation timer is started and the cumulative decrease in pressure time series data is calculated; When the disturbance observation timer reaches the preset observation duration, if the decrease per unit time falls back below the first threshold and the cumulative decrease is less than the preset peeling tolerance equivalent, then the operation of the limit confirmation timer is resumed. If the cumulative drop during the observation period is greater than or equal to the preset peeling tolerance, the limit confirmation timer is reset and the nozzle is determined to re-enter the continuous effective cleaning phase.

[0013] Optionally, during the execution of the limit confirmation timer, the method further includes: Acquire the output power data of the power source driving the flow of cleaning solvent within the nozzle; Calculate the correlation between the synchronous fluctuation of the power source output power data and the pressure time series data; If the correlation degree of synchronous fluctuation is greater than the preset correlation threshold, and the power source output power data is lower than the preset reference power range, then it is determined that the decrease in pressure timing data is caused by power source fluctuation. The limit confirmation timer is suspended and the power compensation program is executed until the power source output power data recovers to the reference power range.

[0014] Optionally, the steps for performing the cleaning endpoint control action include: When the nozzle is determined to be at the cleaning limit stage, close the solvent flushing valve to stop the injection of cleaning solvent into the nozzle; After a preset delay time, the high-pressure gas purging valve is opened to peel off and blow out the softened residue inside the nozzle.

[0015] Optionally, the steps for opening the high-pressure gas purge valve include: The residue softening assessment index is calculated based on the cumulative duration of the rapid removal phase and the continuously effective removal phase. Based on the residue softening assessment index, a preset purging parameter mapping table is retrieved to determine the target purging pressure and target pulse frequency; The high-pressure gas purging valve is controlled to perform intermittent purging according to the target purging pressure and target pulse frequency.

[0016] Secondly, this application also discloses a nozzle self-cleaning process control system for performing the steps in the above method, including: The data acquisition module is used to acquire the pressure timing data of the nozzle during the cleaning process; The feature extraction module is used to extract trend and stability features from pressure time series data; The status recognition module is used to identify the current cleaning stage of the nozzle based on trend and stability characteristics; The control execution module is used to execute cleaning endpoint control actions when the nozzle is determined to be in the cleaning limit stage.

[0017] In summary, the nozzle self-cleaning process control method and system provided in this application can identify the nozzle cleaning stage in real time and dynamically by acquiring pressure time-series data during the nozzle cleaning process and extracting its trend and stability characteristics. Compared with the existing technology that relies solely on a fixed back pressure threshold for judgment, this application can effectively distinguish the cleaning state. Especially when facing complex blockages caused by a mixture of viscous clumps and stubborn gels, by introducing a limit confirmation timer, disturbance recognition, and dynamic compensation mechanism, it can accurately determine the degree of residue softening and execute precise endpoint control. This solution not only avoids the risk of forced exit due to timeout but also reduces false alarms, ensuring the thoroughness of nozzle cleaning, thereby improving the stability and production efficiency of the precision coating process. Attached Figure Description

[0018] Figure 1 This is a schematic flowchart of a nozzle self-cleaning process control method provided in an embodiment of this application.

[0019] Figure 2 This is a schematic diagram of a nozzle self-cleaning process control system provided in an embodiment of this application.

[0020] Figure 3 This is a schematic diagram of the abnormal stall handling process provided in the embodiments of this application.

[0021] Labeling Explanation: 210, Data Acquisition Module; 220, Feature Extraction Module; 230, Status Recognition Module; 240, Control Execution Module. Detailed Implementation

[0022] The technical solutions of this application will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this application, and not all embodiments. The components of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0023] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this application, terms such as "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0024] In precision coating processes, especially in electronic manufacturing dispensing, automatic hot solvent rinsing and purging of the coating nozzles are typically performed between batches to maintain nozzle flow rate and adhesive breakage performance. Due to the heat-sensitive nature of UV adhesives, nozzles are prone to developing residues of varying shapes and bond strengths during heating and prolonged operation cycles. When nozzles continuously process adhesives at different aging stages, the blockages often present a complex state of coexisting viscous clumps and stubborn gels. During solvent rinsing, the outer viscous clumps dissolve first, causing a rapid and significant drop in back pressure inside the nozzle. However, when faced with the slowly softening stubborn gel in the inner layer, the back pressure drop gradient decreases abruptly, slowly approaching a limit value slightly above the ideal cleanliness level. Existing control logic usually relies on a preset fixed back pressure threshold to determine whether cleaning is complete. This approach cannot distinguish between two distinct physical states: basic cleaning with stubborn residue and still effective cleaning. Ultimately, this often leads to incorrect choices between timeout failure and forced exit with residue, severely impacting the continuity and reliability of unmanned automated production processes.

[0025] In this regard, firstly, referring to Figure 1 and Figure 3 This application proposes a nozzle self-cleaning process control method, the method comprising: S1. Acquire pressure timing data of the nozzle during the cleaning process; S2. Extract the trend and stability characteristics of pressure time series data; S3. Identify the current nozzle cleaning stage based on trend and stability characteristics; S4. When the nozzle is determined to be in the cleaning limit stage, execute the cleaning endpoint control action.

[0026] Pressure time series data refers to the set of fluid pressure values ​​collected continuously in chronological order inside or at the nozzle during the nozzle cleaning process. This data can dynamically reflect the real-time resistance changes encountered by the cleaning solvent as it flows inside the nozzle.

[0027] The cleaning limit stage refers to a specific physical state that occurs during the removal of complex blockages. In this state, easily removable residues have been removed, and the remaining stubborn residues soften and peel off extremely slowly under the current solvent flushing conditions. This causes the downward trend of fluid resistance (i.e., pressure) to stagnate, and continuing the current flushing action can no longer bring significant cleaning benefits.

[0028] Regarding the core technical means in the above solution, the following example can be used for implementation: When acquiring pressure timing data of the nozzle during the cleaning process, this can be achieved by installing a standard piezoresistive analog pressure transmitter on the solvent inlet pipe of the nozzle. This pressure transmitter is preferably installed near the upstream of the nozzle and in the main solvent flow channel, so that the measured pressure can more directly reflect changes in the internal resistance of the nozzle, rather than being overly affected by the volume of the distal pipeline, elbow damping, or branch flow diversion. If the nozzle structure allows, the pressure acquisition point can also be set near the nozzle inlet body; if structural limitations prevent close installation, a sampling interface can be set on a straight pipe section connected to the nozzle without any obvious throttling components in between.

[0029] The control system can employ a PLC, industrial controller, or embedded control unit with analog signal acquisition capabilities. Its analog input module reads the 4-20mA current signal output by the transmitter and converts it into the corresponding pressure physical quantity. The pressure unit can be MPa or kPa, but should remain consistent within the same system. To ensure the usability of the conversion results, range calibration can be performed before operation. This involves mapping the lower limit current of the pressure transmitter to the lower limit pressure of the range, and the upper limit current to the upper limit pressure of the range, and writing the corresponding linear conversion relationship into the control program. The acquired pressure values ​​are stored sequentially in the system memory's continuous address blocks, circular buffer, or historical data queue, forming basic data records. To avoid false sampling caused by valve actions or pump start-up shocks at startup, the start time of pressure recording can be set to a short delay after the solvent supply action begins, followed by a stabilization period. Alternatively, after recording begins, the initial segment of data clearly belonging to startup disturbances can be marked as a transition segment and excluded from stage identification.

[0030] When extracting trend and stability features from pressure time-series data, a simple endpoint operation can be used. Specifically, a recent continuous period of pressure time-series data can be used as an analysis period. The length of this analysis period can be set according to the nozzle volume, pipeline response speed, and total cleaning process duration. If the cleaning process changes rapidly, the analysis period can be shortened appropriately; if the cleaning process changes slowly, the analysis period can be extended appropriately. The pressure value at the last time point of this period is subtracted from the pressure value at the first time point, and the difference is used as the trend feature. If the difference is significantly negative, it indicates that the overall pressure has decreased during this period; if the difference is close to zero, it indicates that the pressure change is stagnating; if the difference shows a reverse change, it may mean that local blockages have re-moved, the flow path has been disturbed, or the fluid supply status has changed, requiring further judgment in conjunction with stability features.

[0031] Simultaneously, all pressure values ​​within the analysis period are iterated to identify the maximum and minimum values, and the range between them is calculated. This range is then used as a stability characteristic. A larger range typically indicates more significant pressure fluctuations within the period, potentially corresponding to blockage removal, localized changes in flow channel opening and closing, or strong pump pulsation effects; a smaller range typically indicates more stable pressure changes. Furthermore, upon each new sample, the oldest value can be automatically discarded, and the latest value can be added to the analysis period, thus forming a rolling update of trend and stability characteristics.

[0032] When identifying the current cleaning stage of a nozzle based on trend and stability features, a two-dimensional stage mapping table can be pre-constructed internally. The horizontal axis of this mapping table represents the trend feature interval, and the vertical axis represents the stability feature interval. Different intersection areas correspond to different cleaning stage labels. The pre-constructed stage mapping table can be derived from actual cleaning samples. Specifically, several representative nozzle cleaning tasks can be selected, covering light, moderate, and heavy clogging, as well as varying degrees of glue residue aging. In each task, pressure time-series data, cleaning duration, nozzle state after cleaning, and subsequent test spray results are recorded simultaneously. Then, the trend and stability features from these historical tasks are extracted chronologically and combined with the observed cleaning state to assign a stage label to each feature combination. For example, a period of rapid pressure drop and significant fluctuation can be marked as a rapid cleaning stage; a period where pressure continues to drop but the rate of drop slows and fluctuations begin to converge can be marked as a transitional cleaning stage; and a period where pressure changes are minimal and fluctuations remain low can be marked as a cleaning limit stage. After processing multiple batches of samples, similar feature combinations can be grouped into several intervals to form the mapping table.

[0033] The calculated difference and range are then input into the mapping table. By looking up the table and comparing the results, the current cleaning stage of the nozzle can be output. To avoid misjudgment due to occasional disturbances in a single table lookup, a continuous confirmation mechanism can be set up. That is, the stage label is only officially output when several consecutive rolling analysis cycles fall within the corresponding area of ​​the cleaning limit stage; if a jump occurs in the middle, the observation state continues. This can improve the stability of stage identification.

[0034] When the nozzle is determined to be in the cleaning limit stage, the cleaning endpoint control action, upon receiving the cleaning limit stage tag, immediately sends a digital control command to the actuator, such as cutting off the operation of the solvent supply pump, and simultaneously controls the electromagnetic isolation valve on the solvent pipeline to switch to the closed state, thereby ending the current rinsing process. To ensure a reasonable sequence of actions, the endpoint control action can be executed in the order of stopping the liquid supply, closing the valve, recording the end status, and entering the next process. Recording the end status can include saving the pressure value at the endpoint, trend characteristics, stability characteristics, total cleaning time, and stage tag for subsequent traceability and parameter optimization; entering the next process can be purging, standby, test spray verification, or directly switching to production. If the equipment process requires a short purging after stopping the solvent, the solvent rinsing can be ended first, and then the preset purging sub-process can be initiated, instead of immediately ending all actions after identifying the cleaning limit stage.

[0035] Through the aforementioned overall technical solution, this application overcomes the limitations of traditional endpoint determination based on a fixed absolute pressure threshold. By employing a combination of trend and stability characteristics, it can accurately capture the dynamic evolution of changes in fluid resistance within the nozzle. When the viscous flocs are completely removed, leaving only stubborn gel that softens slowly, the pressure, although not dropping to the traditional extremely low threshold, will show a significantly slower downward trend and its fluctuations will become more stable. This solution can accurately capture this dynamic characteristic through feature recognition, thereby precisely determining the arrival of the cleaning limit stage.

[0036] In some preferred embodiments, the step of acquiring pressure timing data of the nozzle during the cleaning process includes: Extract the original discrete-time sequence of pressure from the nozzle at a preset sampling frequency; The original discrete-time pressure series is validated to remove outliers, resulting in a validated pressure series. The pressure sequence after verification is subjected to sliding window averaging filtering and real-time zero-point compensation to obtain pressure time series data.

[0037] This specific implementation scheme first sets a relatively high sampling frequency to acquire high-density raw pressure discrete-time series data through high-speed acquisition. "Relatively high" here refers to the ability to resolve changes required for subsequent stages of identification; in principle, it should be able to cover sampling requirements such as pump pulsation, residual vibrations during valve switching, and minute changes in nozzle internal resistance. If the sampling frequency is too low, the interval between adjacent data points will be too large, easily missing short-term disturbances and weakening the accuracy of trend judgment; if the sampling frequency is too high, it will increase the computational and storage burden on the controller. Therefore, it can be set in combination with the controller performance and the dynamic response on site.

[0038] Subsequently, a verification mechanism is introduced, setting a reasonable pressure range that conforms to physical laws. The lower limit of this range can be determined based on the allowable fluctuation range near the sensor's zero point, while the upper limit can be determined based on the design pressure-bearing capacity, pumping capacity, and the highest pressure that may occur under normal operating conditions of the cleaning circuit. Abrupt data points exceeding this range are identified as outliers and removed. Removal can be implemented in various ways, such as directly deleting the point, replacing it with the previous valid value, replacing it with the smoothed result of adjacent valid values, or marking the point as invalid and skipping it in subsequent filtering. To avoid mistakenly deleting real rapid pressure changes, outlier determination can also consider the continuity of changes between adjacent sampling points; that is, only when a data point both exceeds the reasonable pressure range and is significantly discontinuous with the preceding and following data points is it considered an outlier.

[0039] Next, a sliding window containing a preset number of data points is constructed. As new data is added, the window slides forward and calculates the arithmetic mean of the data within the window, thus smoothing out the high-frequency mechanical pulsations caused by the liquid supply pump. The window length should be matched with the sampling frequency so that it can cover one or more typical pulsation cycles without excessively flattening the actual pressure drop trend. Simultaneously, reference zero-point data at ambient atmospheric pressure or under pipeline resting conditions can be read in real time. The value after sliding average is subtracted from this reference zero point to complete real-time zero-point compensation. If a gauge pressure transmitter is used, zero-point compensation can mainly be used to correct sensor zero drift; if an absolute pressure transmitter is used, zero-point compensation can also be used to eliminate the influence of ambient pressure changes on the measurement results. Reference zero-point data can be collected once before each cleaning task in a state of no flow and no pressurization, or it can be periodically updated during equipment standby, and the most recent valid reference value can be used during the cleaning process.

[0040] Furthermore, trend characteristics include the decrease per unit time and the rate of decrease decay, while stability characteristics include pressure sliding variance. The steps for extracting trend and stability features from stress time series data include: Linear regression was performed on the pressure time series data within a preset short time window, and the slope of the fitted line was extracted as the decrease per unit time. Calculate the difference or ratio between the unit time decrease of the current short time window and the unit time decrease of the adjacent previous short time window to obtain the rate of decrease decay. The variance of pressure time series data within a preset long-term window is calculated to obtain the pressure sliding variance.

[0041] The rate of decrease per unit time can be understood as the average speed at which the overall pressure changes in a downward direction over a short, continuous period of time.

[0042] The rate of decline decay can be understood as the ability to decline within the current short period of time, whether it remains the same, weakens, or significantly weakens compared to the previous short period of time.

[0043] Pressure sliding variance describes the degree of dispersion in pressure fluctuations around its average level over a longer observation period.

[0044] In this specific implementation, the time window is divided into a short time window and a long time window. The short time window is used to capture the immediate trend of pressure decline, while the long time window is used to observe whether the pressure fluctuations have converged overall. The length of the short time window should be sufficient to include multiple consecutive sampling points so that the fitting result is not dominated by a single point; the length of the long time window should be significantly longer than the short time window to assess stability over a wider time range. Within the short time window, instead of relying on the first and last two data points, a linear regression fitting method is used to estimate the overall trend of all pressure data points within the window. In practice, each sampling moment within the window can be paired with the corresponding pressure value and input into the fitting module. The fitting module outputs a straight line that best represents the overall direction of change, and the slope of this line is defined as the decrease per unit time. If the slope is significantly negative, it indicates that the pressure is still decreasing within the current window; if the slope is close to zero, it indicates that the decrease has slowed significantly; if the slope changes positively, it indicates that the pressure may rebound or there may be local disturbances.

[0045] Subsequently, the slope of the current short-time window is compared with the slope of the previous short-time window. This comparison can be done using either a difference method or a ratio method. The difference method is more suitable for observing how much the downward pressure has decreased; the ratio method is more suitable for observing how much downward pressure has been retained. Regardless of the method used, the goal is to obtain the rate of decline decay, that is, to determine whether the downward pressure trend is weakening and to what extent. Two adjacent short-time windows can be updated using a partially overlapping rolling method, which makes the decay assessment smoother.

[0046] Meanwhile, within a long-term window, the dispersion of all pressure data points relative to the average value of that window is calculated, and this dispersion is used as the pressure sliding variance. A large pressure sliding variance indicates that there are still significant pressure fluctuations over a longer observation range; a consistently small pressure sliding variance indicates that the pressure has entered a relatively stable state. For ease of engineering implementation, linear regression, difference or ratio comparisons, and variance calculations can all be performed by function blocks in the PLC, host computer scripts, or data processing modules in the industrial controller. The processing results are written to the stage identification module in real-time as variables for invocation.

[0047] In some preferred embodiments, the cleaning phase includes a rapid cleaning phase, a continuous and effective cleaning phase, an abnormal stagnation phase, and a cleaning limit phase; Based on trend and stability characteristics, the steps to identify the current nozzle cleaning stage include: When the decrease per unit time is between the preset first gradient threshold and the preset second gradient threshold, and the pressure sliding variance is less than the preset first stable threshold, the nozzle is determined to be in the continuous effective cleaning stage. When the absolute pressure value corresponding to the pressure time series data exceeds the preset acceptable residual tolerance zone and the decrease per unit time approaches zero; or when the pressure sliding variance exceeds the preset second stability threshold, the nozzle is determined to be in an abnormal stagnation stage. When the unit time drop is less than the preset first threshold, the drop rate decay is within the preset fluctuation range, and the absolute pressure value corresponding to the pressure time series data falls into the preset acceptable residual tolerance zone, the preset limit confirmation timer is started; during the operation of the limit confirmation timer, if the unit time drop and pressure sliding variance are continuously lower than the second threshold within the preset time window, the nozzle is determined to be in the cleaning limit stage. Otherwise, the nozzle is determined to be in the rapid cleaning phase.

[0048] Specifically, the aforementioned preset first gradient threshold, second gradient threshold, first stable threshold, second stable threshold, first threshold, second threshold, acceptable residual tolerance band, and preset fluctuation range can all be calibrated during the equipment commissioning phase based on the nozzle model, UV adhesive type, cleaning solvent type, rinsing flow rate, and target cleaning standard. Specifically, nozzles actually used on the target production line can be selected, and standardized cleaning can be performed under four conditions: new nozzles, lightly clogged nozzles, moderately clogged nozzles, and heavily clogged nozzles. The pressure changes, post-disassembly residue, and cleaning compliance conclusions at each moment throughout the process can be recorded. These historical records are then archived, and characteristic intervals that can stably distinguish the four cleaning stages are selected and finally written into the parameter table. At the information processing level, the control system can perform stage identification by reading sensor data at fixed intervals. For example, at each sampling moment, the current pressure value is read and combined with historical pressure values ​​within the previous time window to form a set of analytical data.

[0049] The process begins by determining whether the absolute pressure remains high, followed by assessing the magnitude of the pressure drop per unit time and whether it approaches zero. Finally, the pressure sliding variance is considered to determine the stage. If the current pressure continues to decrease and fluctuates smoothly, the cleaning reaction is still progressing effectively, entering the continuous and effective cleaning stage. If the current pressure is significantly high and the decrease has almost stopped, or if the pressure is not high but fluctuates violently, abnormal blockage or turbulence has occurred within the flow channel, entering the abnormal stagnation stage. If the pressure has entered the acceptable residual tolerance zone, and the decrease is small, the change is slow, and the fluctuation is weak, the cleaning process is approaching its physical limit, and a limit confirmation timer is activated for further confirmation. Other cases that do not fall into the above three specific conditions are uniformly classified into the rapid cleaning stage. This sequential screening approach ensures clear stage determination logic, stable execution, and facilitates the implementation of the controller using a program flow.

[0050] This specific implementation plan defines four cleaning stages with clear physical meanings and provides a strict logic for determining the combination of features. The continuous and effective cleaning stage corresponds to the process where the main blockage inside the nozzle is being steadily and continuously dissolved. At this time, the pressure exhibits a stable and obvious decreasing gradient, and the flow field does not fluctuate drastically. The abnormal stagnation stage corresponds to the cleaning process encountering abnormal obstacles, such as large semi-cured clumps completely blocking the flow channel, or stubborn rubber tumbling within the flow channel causing extreme turbulence in the flow field. The cleaning limit stage corresponds to the removal of easily removable material, leaving only the extremely difficult-to-dissolve bottom layer gel. At this time, the pressure drop is extremely weak and tends to be stable. The rapid cleaning stage usually occurs in the early stages of rinsing, where a large amount of loose residue is quickly washed away, and the pressure exhibits a drastic and rapid decrease. As a fallback logic, when the features do not meet the other three specific stages, they are classified into this stage.

[0051] In practical implementation, a series of threshold parameters can be preset internally. For example, the first and second gradient thresholds are used to define the effective and stable rate of descent range; the acceptable residual tolerance zone is an absolute pressure range used to initially determine whether the pressure has dropped to a reasonably clean range. When the rate of descent per unit time is detected to be extremely small (approaching zero) and the absolute pressure is still very high (outside the tolerance zone), it indicates that the flushing action has failed to reduce any resistance, i.e., a blockage has occurred; or when the pressure slip variance is extremely large (exceeding the second stability threshold), it indicates that there is severe internal fluid disturbance, both of which are judged as abnormal stagnation. When determining the cleanliness limit stage, not only is it required that the rate of descent per unit time be extremely small (less than the first threshold), the decay rate be stable, and the absolute pressure fall into the tolerance zone, but more importantly, a limit confirmation timer is introduced. Only when these small changes are maintained continuously within the preset time window of the timer (i.e., the rate of descent and variance are continuously below the more stringent second threshold) is the entry into the cleanliness limit stage finally confirmed.

[0052] More specifically, the aforementioned threshold parameters can be obtained through prototype calibration and mass production correction. In the prototype calibration stage, multiple nozzles are selected under experimental conditions to prepare samples with different levels of residue and different blockage morphologies, such as outer layer soft rubber residue samples, inner layer gel residue samples, large semi-cured rubber clump blockage samples, and stubborn rubber rollover samples. Then, the same set of cleaning solvents, the same ambient temperature range, and the same pump operating mode are used for rinsing, and the pressure curves for each type of sample are recorded at each cleaning stage. After cleaning, the nozzles are disassembled and inspected, and the pressure curve characteristics are correlated with the disassembly and inspection results to determine the approximate range of each threshold. In the mass production correction stage, these initial thresholds are then written into the actual equipment. Real production data is collected through a trial operation cycle, and a few misjudged samples are reviewed. The corresponding thresholds are appropriately tightened or relaxed until the stage identification is stable.

[0053] The acceptable residual tolerance zone indicates that the pressure state is close to the acceptable cleanliness level allowed by the product process. This tolerance zone can be established as follows: First, select a batch of nozzles that have been manually confirmed to meet the requirements for re-coating, and record their absolute pressure range before and after the cleaning termination; then select a batch of nozzles that, although with lower pressure, still have residues affecting coating quality, and record their absolute pressure performance; compare the two and define a pressure range that can cover qualified samples while minimizing the exclusion of unqualified samples, as the acceptable residual tolerance zone. The purpose of the limit confirmation timer is to prevent hasty conclusions about the cleaning endpoint based solely on a single instantaneous sampling point. In actual cleaning, even if the nozzle is nearly clean, factors such as local fluid disturbances, pump pulsations, and sensor noise can still cause short-term fluctuations at individual points. Without a time window for confirmation, cleaning may end prematurely at a moment of accidental stability. With the introduction of the limit confirmation timer, a state of low pressure drop, low fluctuation, and pressure falling within the tolerance zone is required to be maintained for a preset time, indicating that no significant changes with cleaning value are occurring inside the nozzle, before entering the cleaning limit stage. This design effectively improves the reliability of endpoint determination.

[0054] In some preferred embodiments, when it is determined that the nozzle is in an abnormal stagnation phase, the method further includes: If the absolute pressure value exceeds the acceptable residual tolerance zone and the pressure drop per unit time is zero, the liquid supply pump will stop and a fault signal will be output. If the pressure sliding variance fluctuates and the total flushing time reaches the preset safe duration threshold, an alternating gas-liquid impact action is executed, and the cleaning process is determined based on the pressure sliding variance after the impact to either continue cleaning or issue a cleanliness doubt warning.

[0055] In practice, stopping the liquid supply pump and outputting a fault signal can be directly accomplished by the control system via relays, driver enable signals, or bus control commands. The fault signal can include at least one local output and at least one remote output, such as illuminating a red alarm light, triggering a buzzer, sending a fault code to the host computer, or displaying a message on the human-machine interface indicating nozzle blockage and requesting manual intervention. This setup aims to prevent the equipment from continuing to operate under high load and wasting energy when absolute blockage is confirmed, instead quickly entering a controlled shutdown state. Performing the alternating gas-liquid impact only after the total flushing time reaches a preset safe duration threshold is to avoid prematurely entering a strong intervention mode when only short-term fluctuations occur. This is because some slight fluctuations may only be a transitional state before large pieces of residue are about to detach; immediately applying a strong impact at this time would only increase the burden on the cleaning system.

[0056] The preset safe time threshold can be determined experimentally. Essentially, it is the judgment boundary where, even after sufficient time has been given for regular rinsing, there is still no improvement. This threshold can be determined by statistically analyzing the approximate time distribution of a normal nozzle from the start of cleaning to reaching the cleaning limit stage under standard processes, and then combining this with the heat load allowed for continuous rinsing by the equipment, solvent consumption, and pump tolerance to select a reasonable time as the safe time threshold.

[0057] The pressure sliding variance after the impact is used to determine whether to continue cleaning or issue a cleanliness doubt warning. This is because the pressure sliding variance can directly reflect whether the internal turbulent flow field has been effectively eliminated. If the variance decreases significantly after the impact and the subsequent pressure curve returns to a stable decline, it indicates that the previously turbulent residue has been broken up or carried away, and the normal cleaning process can be resumed. If the variance continues to fluctuate violently, it indicates that the state of the residue has not improved or that complex blockages that are difficult to remove online have formed inside the nozzle. In this case, continuing automatic cleaning is of little value, and a cleanliness doubt warning should be issued to the operator.

[0058] This specific implementation plan designs differentiated handling strategies for two different causes of abnormal stagnation. The first scenario is when the pressure remains high without any downward trend. In this case, the primary task is to protect the hardware; therefore, the power to the liquid supply pump is directly cut off, and a fault signal is sent to the host computer or on-site indicator lights to call for manual intervention, preventing the equipment from operating with a malfunction. The second scenario is when the overall pressure level may not be high, but the variance fluctuates drastically, indicating that there are stubborn residues inside the nozzle that move with the fluid but cannot be dislodged. In this case, if the flushing time has reached the safe duration threshold (meaning that conventional flushing has been fully attempted but ineffective), the physical state of the cleaning medium will be actively changed, triggering an alternating gas-liquid impact action.

[0059] To more fully explain the identification and handling of the second scenario, it can be understood as a state where residual material is not completely blocked but cannot be completely removed. For example, one end of a sheet of rubber adheres to the inner wall of the nozzle, while the other end oscillates back and forth with the fluid. It may momentarily block the flow channel and at other moments be pushed open by the fluid, causing significant pressure fluctuations. In this state, simply extending the constant flushing is often ineffective because the flow field lacks the instantaneous mechanical disturbance to break the rubber adhesion point or tear the semi-cured structure. Alternating gas-liquid impact is designed specifically to address this physical problem. Its essence is not to continue increasing the normal flow rate, but to artificially create periodic media state switching and impact fluctuations, thereby improving the ability to remove stubborn residual material. In contrast, the first scenario involves almost no effective flow through the blocked area. Implementing alternating gas-liquid impact in this case may not be safe, as if upstream pressure continues to accumulate while downstream pressure cannot be released, it increases the risk of pressure buildup. Therefore, this solution first implements pump shutdown protection for solid blockages, and then applies impact drying to dynamically stuck areas.

[0060] In one specific embodiment, the alternating gas-liquid impact action can be achieved by a control system coordinating the high-frequency switching of the solvent valve and the high-pressure gas valve. For example, while controlling the solvent pump to continuously supply liquid, the high-pressure gas solenoid valve connected to the flushing pipeline is rapidly opened and closed at a frequency of 2Hz to 5Hz. The instantaneous injection of high-pressure gas creates strong cavitation and water hammer effects in the liquid solvent. This high-frequency mechanical impact force can effectively shatter or peel off semi-cured UV adhesives that are immune to pure chemical dissolution.

[0061] After performing an impact for a preset time (e.g., 5 seconds), reassess the pressure slip variance. If the variance converges significantly, it indicates that the stubborn rubber has been broken up and expelled, and the normal cleaning logic can be resumed; if the variance still fluctuates wildly, it indicates that the impact is ineffective, and a cleanliness doubt warning is output, suggesting that the nozzle may need offline ultrasonic deep cleaning.

[0062] Furthermore, the hardware can include a solvent passage, a high-pressure gas passage, a one-way anti-backflow structure, and a solenoid valve control unit. The high-pressure gas passage is preferably located near the upstream of the nozzle to shorten the response time of the gas reaching the blockage area after injection; simultaneously, an anti-backflow component is installed before the gas-liquid junction to prevent solvent from flowing back into the gas path. Upon receiving a command to execute alternating gas-liquid impact, the system first confirms that the liquid supply pump is functioning normally, the solvent valve is open, and the gas pressure is within the allowable range, then drives the high-pressure gas valve to periodically open and close at a preset frequency. The aforementioned frequency of 2Hz to 5Hz is suitable for scenarios where rapid pulses create significant disturbances without causing the valve to operate too frequently and become unstable; in cases where the nozzle flow channel is narrower or the residue is more fragile, a more suitable frequency range can be selected through experimentation and fixed as a parameter group corresponding to different nozzle models.

[0063] In terms of the processing flow, the entire impact action can be divided into a preparation phase, an impact phase, and an evaluation phase. The preparation phase checks whether safety interlock conditions are met, such as confirming there is no overpressure alarm, that the gas source is normal, and that the drainage path is unobstructed. The impact phase injects gas pulses into the liquid flow according to a preset rhythm for a preset duration. The evaluation phase continues to collect pressure time-series data after the impact, comparing the pressure slip variance, the decrease per unit time, and the absolute pressure change before and after the impact. If the pressure curve changes from its previous chaotic and violent fluctuations to a gradual decrease, it indicates that the impact has indeed broken the original stuck structure; if there is almost no improvement before and after the impact, it suggests that the online automatic cleaning method has reached its limit.

[0064] In some preferred embodiments, during the operation of the limit confirmation timer, the method further includes: If the absolute value of the decrease per unit time is detected to be greater than the first threshold, the limit confirmation timer is suspended and the disturbance identification state is entered. In the disturbance identification state, a preset disturbance observation timer is started and the cumulative decrease in pressure time series data is calculated; When the disturbance observation timer reaches the preset observation duration, if the decrease per unit time falls back below the first threshold and the cumulative decrease is less than the preset peeling tolerance equivalent, then the operation of the limit confirmation timer is resumed. If the cumulative drop during the observation period is greater than or equal to the preset peeling tolerance, the limit confirmation timer is reset and the nozzle is determined to re-enter the continuous effective cleaning phase.

[0065] The aforementioned suspended limit confirmation timer is preferably understood as pausing the timer but retaining the currently accumulated confirmation duration, rather than directly resetting it to zero. The control program can set a timer status flag, including at least three states: running, suspended, and reset / restart. When the absolute value of the decrease per unit time is detected to be greater than the first threshold, the timer state is switched from running to suspended, the currently accumulated time is latched, and an independent disturbance observation timer is started.

[0066] The cumulative decrease in pressure time-series data can be achieved by continuously recording pressure changes during the operation of the disturbance observation timer. Specifically, the following method can be used: using the pressure value at the start of the disturbance identification state as the observation starting point, continuously collecting pressure values ​​over the subsequent observation period, and accumulating the total actual decrease during this time. The resulting cumulative decrease can reflect whether the disturbance was merely a small, instantaneous oscillation or truly resulted in a sustained release of resistance with cleanliness implications.

[0067] The preset peeling tolerance equivalent can be obtained experimentally. Specifically, multiple nozzles that are close to their cleaning limits can be selected. During the limit confirmation phase, several minor disturbance events are manually recorded. After cleaning, the nozzles are disassembled and the corresponding events are checked to see if there is significant residual layer peeling. If some disturbance events cause a temporary drop in pressure, but after disassembly it is found that only a small amount of debris falls off and does not affect the overall residual layer structure, then the cumulative drop corresponding to these events is classified as within the tolerable range. If some events correspond to significant exposure of new layers and subsequent continuous effective cleaning, then their cumulative drop is used as the reference lower limit for triggering a return to the continuous effective cleaning phase. Through this summary, a range value suitable for the current nozzle and adhesive system can be set for the peeling tolerance equivalent, and finally written into the control parameters.

[0068] In the specific implementation plan, when a sudden increase in the rate of pressure drop (absolute value exceeding the first threshold) is detected during the limit confirmation period, the previous limit confirmation process is not immediately rejected. Instead, the limit confirmation timer is suspended (i.e., the timing is paused but the currently accumulated time is retained), and a dedicated disturbance identification state is entered. In this state, a new disturbance observation timer is started, and the cumulative pressure drop during this period is calculated by integration. This cumulative pressure drop represents the total actual reduction in resistance inside the nozzle during this disturbance, indirectly reflecting the volume of the flaking residue.

[0069] To further explain this buffering mechanism, consider this scenario: When the limit confirmation timer has been running for some time, indicating the nozzle is considered close to the cleaning endpoint, a short-term pressure drop could indicate the shedding of a tiny piece of debris or the removal of a residual piece of actual thickness. Directly resetting the limit confirmation timer would be overly sensitive; ignoring it completely could miss the opportunity to re-enter the effective cleaning phase. By setting a disturbance recognition state, the original endpoint confirmation is paused, and then the disturbance is observed within a short observation window to see if it ends quickly or only causes a small total pressure drop. If so, the disturbance is insufficient to change the overall judgment that the nozzle is nearing its limit; if not, the internal cleaning conditions of the nozzle have substantially changed, and the process should revert to the continuous effective cleaning phase. This buffering mechanism, employing disturbance observation and equivalent assessment, balances sensitivity and stability.

[0070] In practice, a peeling tolerance threshold can be preset, which can be a very small pressure difference (e.g., 0.02 MPa). After the disturbance observation timer completes the preset observation period (e.g., 3 seconds), a judgment is made: if the cumulative drop during this period is very small (less than the peeling tolerance threshold) and the drop rate has subsided (falling back below the first threshold), it is considered to be just an insignificant minor debris shedding or sensor noise, and the previous confirmation time continues to accumulate as the limit confirmation timer resumes operation; conversely, if the cumulative drop reaches the peeling tolerance threshold, it indicates that a significant piece of residue has been peeled off, exposing a new layer to be cleaned. At this time, the limit confirmation timer must be completely reset to zero, and the nozzle is determined to return to the continuous effective cleaning phase, restarting a new round of cleaning and confirmation cycle. The above 0.02 MPa and 3 seconds are only implementation examples, and their applicability is premised on the rated pressure range of the nozzle flushing circuit, the sensor resolution, and the typical pressure drop caused by residue peeling being matched with this level. In practical deployments, if the equipment uses a larger pressure range, a larger loop volume, or different nozzle flow channel dimensions, the spalling tolerance and observation time can be set using code parameters and obtained through experimental calibration. For example, multiple real disturbance events can be continuously collected under near-cleaning limits, and the disassembly and inspection results after each disturbance and the subsequent cleaning time changes can be compared. Then, a parameter combination that can tolerate minor and harmless disturbances and promptly identify effective spalling can be selected.

[0071] In some preferred embodiments, during the operation of the limit confirmation timer, the method further includes: Acquire the output power data of the power source driving the flow of cleaning solvent within the nozzle; Calculate the correlation between the synchronous fluctuation of the power source output power data and the pressure time series data; If the correlation degree of synchronous fluctuation is greater than the preset correlation threshold, and the power source output power data is lower than the preset reference power range, then it is determined that the decrease in pressure timing data is caused by power source fluctuation. The limit confirmation timer is suspended and the power compensation program is executed until the power source output power data recovers to the reference power range.

[0072] Synchronous fluctuation correlation can be understood as whether the changing trends of two curves are consistent and whether their fluctuations are synchronized within the same time window. For example, when the output power of the power source decreases, the pressure time series data also decreases synchronously, and the start time, duration, and recovery time of the decrease are basically the same, it indicates that there is a high degree of synchronicity between the two. Conversely, if the power source is basically stable, but the pressure changes alone, it is more likely to be caused by changes in the internal resistance of the nozzle.

[0073] The preset correlation threshold and preset baseline power range can also be established through experiments. Specifically, during the equipment commissioning phase, two types of sample operating conditions can be created: one is a condition where real blockages are gradually cleared while the power source remains stable, and the other is a condition where the pump power output is artificially reduced while the internal blockage state of the nozzle remains basically unchanged. The synchronization relationship between the pressure curve and the power source curve under these two operating conditions is then recorded to screen out the correlation threshold that can stably distinguish between changes in internal cleaning and changes in external power. Simultaneously, while the pump is in normal operating condition, the power source output power data is continuously recorded for a period of time to obtain the fluctuation range during stable equipment operation, and this range is set as the baseline power range. When it is determined that the pressure drop is caused by power source fluctuations, the reason for suspending the limit confirmation timer is that the low pressure at this point does not necessarily mean the nozzle is cleaner, but only that the flushing driving force has weakened. If the limit confirmation timer continues to accumulate time, invalid low-pressure flushing errors under non-standard operating conditions will be included in the final confirmation process. Executing the power compensation program is to quickly restore the standard cleaning conditions that can be used to judge the true state of the nozzle.

[0074] In practice, not only is the pressure timing data at the nozzle collected, but the power output data of the liquid supply pump is also collected simultaneously. The power output data can be flexibly obtained according to the type of pump. For example, for a pneumatic diaphragm pump, the inlet pressure or flow rate of the drive air circuit can be collected; for an electric gear pump, the real-time current or torque feedback of the servo motor can be collected.

[0075] Subsequently, mathematical statistical algorithms (such as Pearson correlation coefficient calculation) are used to analyze the correlation of synchronous fluctuations between the two time series within the same time window in real time. In practical implementation, the following approach can be adopted: extract the pressure change sequence and the power source change sequence within the same time window, and compare whether they decrease or rise simultaneously, or whether one changes while the other remains essentially unchanged; if similar directions of change and similar temporal synchronicity are observed at multiple consecutive sampling points, the correlation of synchronous fluctuations is considered high. This processing can be completed by the software module of a PLC, industrial computer, or embedded controller.

[0076] For pneumatic diaphragm pumps, the power source output power data can be obtained by installing a pressure sensor and optional flow monitoring device at the inlet of the drive air circuit to periodically read the drive air pressure and airflow status. When the total air supply in the factory decreases due to external equipment competing for air, these data will change before or simultaneously with the decrease in pump output capacity. For electric gear pumps, the power source output power data can be obtained by directly reading feedback quantities such as real-time current, torque percentage, or speed closed-loop error from the servo driver. When the mains voltage fluctuates or the driver current is limited, the above feedback quantities will show abnormalities.

[0077] Subsequently, during the limit confirmation period, if an abnormal drop in pressure timing data is detected, the power source output power data for the same period will be retrieved immediately using the aforementioned method. If the calculated correlation of synchronous fluctuations is extremely high (e.g., a correlation coefficient greater than the correlation threshold of 0.85), it indicates that the pressure drop trajectory highly overlaps with the power source attenuation trajectory. At this point, the absolute value of the power source is further checked. If it is found that it has indeed fallen below the normal operating baseline power range, it can be determined that the current pressure drop is not due to the nozzle becoming cleaner, but rather due to insufficient pump power. Based on this accurate judgment, the limit confirmation timer is immediately suspended (to prevent the erroneous accumulation of invalid low-pressure flushing time) and the power compensation procedure is triggered. The power compensation procedure may include: automatically increasing the opening of the electro-proportional valve to increase the drive air pressure, or increasing the drive gain of the motor, or even pausing the flushing action and waiting for the plant's air / grid network to recover when the power source is extremely scarce. The suspension is lifted and the normal limit confirmation process resumes only after the power source output power stabilizes again within the baseline power range.

[0078] The correlation coefficient greater than 0.85 mentioned above is only one feasible example used to illustrate the judgment situation of extremely high correlation of synchronization fluctuations. Other calibrated correlation thresholds can also be used under different equipment, different sampling frequencies, and different noise levels. The specific content of the power compensation procedure can be designed according to the power source type. For pneumatic systems, the power compensation procedure may include increasing the target value of the pre-stage pressure regulator, increasing the opening of the proportional valve, temporarily closing non-critical air-consuming branches, switching to the backup air tank, or pausing flushing before the air source is restored. For electric systems, the power compensation procedure may include increasing the allowable output of the drive, temporarily reducing the load of other non-critical actuators, switching to the regulated power supply branch, or waiting for the power grid to stabilize before continuing flushing. In terms of software flow, a power anomaly suspension flag can be set. Once a high synchronization fluctuation is detected and the power source is below the reference power range, the flag is set and the limit confirmation timer is paused; only after the power source output power is confirmed to stabilize again and no longer below the reference power range within a subsequent continuous detection window is the flag cleared and the timer resumed. This design can avoid frequent state switching when the power source has just briefly rebounded and then fallen again, which would lead to unstable judgment.

[0079] In some preferred embodiments, the step of performing the cleaning endpoint control action includes: When the nozzle is determined to be at the cleaning limit stage, close the solvent flushing valve to stop the injection of cleaning solvent into the nozzle; After a preset delay time, the high-pressure gas purging valve is opened to peel off and blow out the softened residue inside the nozzle.

[0080] In practice, once the cleaning limit is confirmed, a shut-off command is immediately sent to the solvent flushing valve to cut off the solvent supply. This action not only avoids the pointless consumption of expensive cleaning solvents but also reduces waste liquid generation. Subsequently, a preset delay time (e.g., 1 to 3 seconds) is initiated. The purpose of this delay time is twofold: firstly, it allows some of the residual solvent inside the nozzle and pipeline to naturally evacuate under gravity or inertia, preventing subsequent high-pressure gas from directly impacting a large amount of liquid solvent and causing violent splashing; secondly, this settling time allows the stubborn residues adhering to the inner wall of the nozzle to fully absorb the residual solvent on the surface, further swelling and softening. After the delay time ends, the high-pressure gas purge valve is opened, injecting dry, clean high-pressure gas (such as compressed air or nitrogen) into the nozzle flow channel.

[0081] To further improve the purging effect and avoid damage to the nozzle, the steps for opening the high-pressure gas purging valve include: The residue softening assessment index is calculated based on the cumulative duration of the rapid removal phase and the continuously effective removal phase. Based on the residue softening assessment index, a preset purging parameter mapping table is retrieved to determine the target purging pressure and target pulse frequency; The high-pressure gas purging valve is controlled to perform intermittent purging according to the target purging pressure and target pulse frequency.

[0082] In actual operation, the thickness and stubbornness of residues inside nozzles vary greatly between different batches. Using fixed parameters for purging may result in insufficient pressure for thorough cleaning or excessive pressure that damages the nozzle's precision micro-orifices. Therefore, the cumulative duration of the rapid cleaning phase and the continuously effective cleaning phase recorded in the preceding process is retrieved first. Longer durations of these two phases generally indicate more initial blockage inside the nozzle and a longer time for the underlying residue to soak and swell in the solvent. A quantified residue softening assessment index is calculated using a preset weighted algorithm (e.g., multiplying the durations of the two phases by their respective weighting coefficients and then summing the results). This index objectively reflects the current physical state of the underlying residue.

[0083] Next, the index is used to retrieve a pre-calibrated purging parameter mapping table from memory. This mapping table was established during the equipment commissioning phase through fitting with a large amount of experimental data, and it internally stores the correspondence between different softening index ranges and optimal purging pressure and pulse frequency. For example, when the softening index is high (the residue is fully soaked and soft), the mapping table will output a lower target purging pressure and a higher target pulse frequency; when the softening index is low (the residue is hard), it will output a higher target purging pressure and a lower target pulse frequency.

[0084] Finally, according to the target parameters obtained from the lookup table, the high-pressure gas purging valve is driven to perform intermittent purging. Intermittent purging refers to controlling the solenoid valve to open and close rapidly at the target pulse frequency, so that the high-pressure gas enters the nozzle in the form of pulse waves. This not only ensures the final cleaning quality, but also maximizes the service life of the precision nozzle.

[0085] Secondly, referring to Figure 2 This application also proposes a nozzle self-cleaning process control system for performing the steps in the above method, including: Data acquisition module 210 is used to acquire pressure timing data of the nozzle during the cleaning process; Feature extraction module 220 is used to extract trend features and stability features of pressure time series data; The status recognition module 230 is used to identify the current cleaning stage of the nozzle based on trend characteristics and stability characteristics; The control execution module 240 is used to execute the cleaning endpoint control action when it is determined that the nozzle is in the cleaning limit stage.

[0086] Through the above technical solution, the system highly integrates complex pressure data acquisition, multi-dimensional feature extraction, dynamic stage identification, and intelligent endpoint control. The data acquisition module 210 ensures high-fidelity acquisition of underlying physical quantities; the feature extraction module 220 transforms raw data into physically meaningful trend and stability indicators; the state identification module 230, combined with multiple decision logics and anti-interference mechanisms, determines the cleaning limit stage; and the control execution module 240 implements the optimal endpoint closing action based on the decision results. The collaborative work of these modules completely breaks through the limitations of traditional methods that determine the cleaning endpoint based on fixed pressure thresholds, endowing the equipment with the ability to autonomously judge, self-protect, and adaptively adjust under complex working conditions, significantly improving the continuity, reliability, and nozzle reuse success rate of the unmanned automated production process.

[0087] The above are merely embodiments of this application and are not intended to limit the scope of protection of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A method for controlling the self-cleaning process of a nozzle, characterized in that, The method includes: Acquire pressure timing data of the nozzle during the cleaning process; Extract the trend and stability features of the pressure time series data; Based on the trend characteristics and the stability characteristics, the current cleaning stage of the nozzle is identified; When the nozzle is determined to be in the cleaning limit stage, the cleaning endpoint control action is executed.

2. The nozzle self-cleaning process control method according to claim 1, characterized in that, The step of acquiring the pressure timing data of the nozzle during the cleaning process includes: The original pressure discrete-time sequence of the nozzle is extracted at a preset sampling frequency; The original discrete-time pressure sequence is validated to remove outliers, resulting in a validated pressure sequence. The verified pressure sequence is subjected to sliding window averaging filtering and real-time zero-point compensation to obtain the pressure time series data.

3. The nozzle self-cleaning process control method according to claim 1, characterized in that, The trend features include the decrease per unit time and the rate of decrease decay, and the stability features include the pressure sliding variance; The steps for extracting the trend and stability features of the pressure time series data include: Linear regression is performed on the pressure time series data within a preset short time window, and the slope of the fitted straight line is extracted as the unit time decrease. Calculate the difference or ratio between the unit time decrease of the current short time window and the unit time decrease of the adjacent previous short time window to obtain the decrease rate attenuation. The variance of the pressure time series data within a preset long time window is calculated to obtain the pressure sliding variance.

4. The nozzle self-cleaning process control method according to claim 3, characterized in that, The cleaning phases include a rapid cleaning phase, a continuous and effective cleaning phase, an abnormal stagnation phase, and a cleaning limit phase. The step of identifying the current cleaning stage of the nozzle based on the trend characteristics and the stability characteristics includes: When the unit time decrease is between a preset first gradient threshold and a preset second gradient threshold, and the pressure sliding variance is less than a preset first stability threshold, the nozzle is determined to be in a continuous effective cleaning phase. When the absolute pressure value corresponding to the pressure time series data exceeds the preset acceptable residual tolerance band, and the decrease per unit time tends to zero; or, when the pressure sliding variance exceeds the preset second stability threshold, the nozzle is determined to be in an abnormal stagnation stage. When the unit time drop is less than a preset first threshold, and the drop rate decay is within a preset fluctuation range, and the absolute pressure value corresponding to the pressure time series data falls into a preset acceptable residual tolerance zone, a preset limit confirmation timer is started; during the operation of the limit confirmation timer, if the unit time drop and the pressure sliding variance are continuously lower than the second threshold within a preset time window, the nozzle is determined to be in the cleaning limit stage. Otherwise, the nozzle is determined to be in the rapid cleaning phase.

5. The nozzle self-cleaning process control method according to claim 4, characterized in that, When it is determined that the nozzle is in an abnormal stagnation phase, the method further includes: If the absolute pressure value exceeds the acceptable residual tolerance zone and the decrease per unit time is zero, the liquid supply pump is stopped and a fault signal is output. If the pressure sliding variance fluctuates and the total flushing time reaches the preset safe duration threshold, then an alternating gas-liquid impact action is performed, and the cleaning process is continued or a cleanliness doubt warning is issued based on the pressure sliding variance after the impact.

6. The nozzle self-cleaning process control method according to claim 4, characterized in that, During the operation of the limit confirmation timer, the method further includes: If the absolute value of the decrease per unit time is detected to be greater than the first threshold, the limit confirmation timer is suspended and the disturbance identification state is entered. In the disturbance identification state, a preset disturbance observation timer is started and the cumulative decrease of the pressure time series data is calculated; When the disturbance observation timer reaches the preset observation duration, if the decrease per unit time falls back below the first threshold and the cumulative decrease is less than the preset peeling tolerance equivalent, then the operation of the limit confirmation timer is resumed. If the cumulative decrease during the observation period is greater than or equal to the preset peeling tolerance equivalent, then the limit confirmation timer is reset and the nozzle is determined to re-enter the continuous effective cleaning phase.

7. The nozzle self-cleaning process control method according to claim 4, characterized in that, During the operation of the limit confirmation timer, the method further includes: Acquire the output power data of the power source driving the flow of cleaning solvent within the nozzle; Calculate the correlation degree of synchronous fluctuation between the power source output power data and the pressure time series data; If the correlation degree of the synchronous fluctuation is greater than the preset correlation threshold, and the power source output power data is lower than the preset reference power range, then it is determined that the decrease in the pressure timing data is caused by the power source fluctuation, the limit confirmation timer is suspended and the power compensation program is executed until the power source output power data recovers to the reference power range.

8. The nozzle self-cleaning process control method according to claim 4, characterized in that, The steps for performing the cleaning endpoint control action include: When the nozzle is determined to be in the cleaning limit stage, the solvent flushing valve is closed to stop the injection of cleaning solvent into the nozzle; After a preset delay time, the high-pressure gas purging valve is opened to peel off and blow out the softened residue inside the nozzle.

9. A nozzle self-cleaning process control method according to claim 8, characterized in that, The step of opening the high-pressure gas purging valve includes: The residue softening assessment index is calculated based on the cumulative duration of the rapid removal phase and the continuous effective removal phase. Based on the residue softening assessment index, a preset purging parameter mapping table is retrieved to determine the target purging pressure and target pulse frequency; The high-pressure gas purging valve is controlled to perform intermittent purging according to the target purging pressure and the target pulse frequency.

10. A nozzle self-cleaning process control system for performing the steps of the method according to any one of claims 1 to 9, characterized in that, include: The data acquisition module is used to acquire the pressure timing data of the nozzle during the cleaning process; The feature extraction module is used to extract the trend features and stability features of the pressure time series data; A status recognition module is used to identify the current cleaning stage of the nozzle based on the trend characteristics and the stability characteristics; The control execution module is used to execute a cleaning endpoint control action when it is determined that the nozzle is in the cleaning limit stage.

Images