Delay method and system for spray head printing software

Abstract

The invention relates to the technical field, in particular to a delay method and system for nozzle printing software, and the method comprises the following steps: carrying out the numbering and classification of real-time data and the item-by-item splitting of start and stop behaviors, and carrying out the stage attribution of the delay change of a nozzle at a microscopic level; a structured time sequence is formed under a unified dimension, the accuracy of subsequent calculation is established, linkage judgment is constructed through response behaviors under different load environments in combination with the ink jet amount and the pressure difference, the stability of starting delay adjustment is guaranteed, image errors are adopted as an external observation channel, a closed-loop control path is established on the time level, and the stability of delay adjustment is improved. Through coupling operation of three paths of real-time track processing, difference matrix generation, image error mapping and period correction, delay control does not depend on fixed parameter or mean value adjustment any more, dynamic mapping and fault isolation capability are achieved, and flexibility, feedback granularity and precision control of delay management are enhanced.

Description

Technical Field

[0001] The present invention relates to the field of printing control technology, and in particular to a time delay method and system for nozzle printing software. Background Art

[0002] Print control technology aims to control the hardware and software in the printing device so that the links in the printing process can run in a coordinated and precise manner. In inkjet printers, controlling the inkjet timing and emission signal of the nozzle is a key technology. The delay control method accurately adjusts the working timing of the nozzle to make the position, size and shape of the ink ejection consistent with the expectation, thereby improving the printing quality. Print control technology not only includes the synchronization of hardware signals, but also covers the control algorithm at the software level, optimizing the printing speed and accuracy.

[0003] The delay method used in print head printing software controls the delay when the print head ejects ink, accurately adjusts the working state of the print head, optimizes the start and stop time of the print head, and makes the ink ejection meet the requirements of the printed image. The purpose is to improve the print quality, reduce errors and inconsistencies in the printing process, achieve high-precision and high-quality printing effects, improve the accuracy of inkjet printing, effectively reduce printing defects caused by improper print head control, and ensure the stability of print quality while achieving higher printing speeds.

[0004] The existing technology only implements parameter scheduling based on fixed delay control and static synchronization strategy during the nozzle control process, and fails to achieve dynamic tracking of delay changes in the entire cycle, resulting in sluggish delay response performance in high-frequency tasks or thermal load scenarios. In addition, the existing technology lacks an abnormality recognition mechanism based on multi-parameter drive. There are problems of abnormal pressure fluctuations or uneven ink jetting in the nozzle in a short period of time, and it is difficult to provide timely feedback and adjustment in the next cycle, resulting in repetitive nozzle offset or image breakage, missing the valuable correction channel in high-precision control, and making the system heavily dependent on empirical parameters for the control of edge alignment and morphological fidelity in complex image printing. In continuous printing tasks, delay tailing and distribution imbalance are easily formed, resulting in quality defects such as ink droplet accumulation and sparse distribution. Summary of the Invention

[0005] The purpose of the present invention is to solve the shortcomings of the prior art and to propose a delay method and system for nozzle printing software.

[0006] In order to achieve the above object, the present invention adopts the following technical solution: a time delay method for nozzle printing software, comprising the following steps: Step 1: Sort the operation history data based on the real-time working data of the sprinkler, identify the delay changes during the operation of the sprinkler, extract the sprinkler timing characteristics and working status characteristics, generate the real-time operation time series of the sprinkler, and construct the sprinkler time series data set based on the time series data; Step 2: Based on the nozzle timing data set, calculate the time difference between the start and stop of the nozzle and calculate the delay optimization objective function, calculate the delay adjustment of the nozzle according to the model, and generate the delay optimization model of the nozzle under different working conditions; Step 3: Based on the delay optimization model, the start-up timing of the printhead under different workloads is obtained and the relationship between the printhead pressure and the ink ejection volume is analyzed to determine whether the printhead is overloaded or abnormal. The timing when the pressure is within the normal range is obtained, and a start-up delay adjustment plan for the printhead is generated; Step 4: Based on the inkjet image captured in real time by the start-up delay adjustment scheme, analyze the morphology and ejection direction of the inkjet printhead, compare the difference between the actual inkjet effect and the expected image, calculate the image error, use the image recognition result to fine-tune the operation timing of the printhead, and generate a printhead timing correction signal; Step 5: Track and optimize based on the nozzle timing correction signal, adjust the nozzle timing in combination with the inkjet feedback data after the printing task, obtain the nozzle operation reference map, start rhythm matching index and historical injection calibration matrix, and generate a nozzle operation optimization model.

[0007] As a further solution of the present invention, the specific steps of generating the nozzle timing data set are: Based on the real-time working data of the sprinkler, the trigger instruction time field and the response feedback time field in the data are intercepted and the time difference is calculated. The sprinkler number field is used to classify each data and number them in time sequence. After arranging them in ascending order by the time field, a mapping table of sprinkler numbers and time differences is established to generate a sprinkler delay interval sequence; Based on the nozzle delay interval sequence, the delay value and the corresponding recorded injection frequency value of the nozzle are extracted and paired, the nozzle temperature data of the temperature change amplitude in the same time period is recorded, the injection event count value is collected and the event density per unit time is calculated, and the three indicators are constructed into a continuous vector sequence to establish the nozzle state feature matrix; Based on the nozzle state feature matrix, the time tags of all nozzle control instructions in the operation sequence are read, the control pulse width and trigger state mark at the time point are collected, the injection confirmation flag related to the state change value is extracted, and the data is aligned and summarized according to the time sequence to obtain the nozzle timing data set.

[0008] As a further solution of the present invention, the specific steps of generating the delay optimization model are: Based on the nozzle timing data set, the numerical difference between the start time field and the stop time field in the record is calculated, the time difference values ​​of the nozzle numbers are classified and sorted, and the data are arranged in order using single numbers and the extreme values ​​and arithmetic mean values ​​of the data are calculated to generate a nozzle operation cycle statistics table; Based on the nozzle operation cycle statistics table, the difference in average operation time is calculated and a number index table is established. The period item corresponding to the peak deviation number is extracted and the absolute value of the deviation is calculated. The interval range is established by combining the valley period and the peak period and forming a difference matrix to establish a nozzle delay difference vector set; Based on the nozzle delay difference vector set, the long short-term memory network is used to extract the workload value under the nozzle number and perform normalized ratio conversion with the period deviation value. The ratio sets of the same number under different temperature conditions and trigger frequencies are sorted out, and the numbers are divided into state categories and filled with corresponding values ​​in each state to obtain the nozzle delay optimization model.

[0009] As a further solution of the present invention, the long short-term memory network is according to the formula: ; in: is the normalized ratio after nozzle delay optimization, is the attention weight coefficient of the numbered nozzle at time t, is the workload prediction value output by the long short-term memory network at time t for the sprinkler number, is the load correction weight coefficient of the ambient temperature parameter T, where T is the working ambient temperature of the current nozzle number. is the load correction weight coefficient of the trigger frequency parameter f, f is the current nozzle trigger frequency, is the delay fluctuation parameter The load correction weight coefficient, is the maximum delay time difference of the nozzle in the past three cycles, is the cooling rate parameter The load correction weight coefficient, is the cooling rate of the nozzle temperature drop, is the load standard deviation The penalty adjustment coefficient, is the workload standard deviation of the sprinkler number at time t, is the periodic deviation value of the nozzle number at time t, is the period deviation zero value compensation constant, is the modulation coefficient of the exponential deviation correction term, is the natural exponential function, is the attenuation coefficient in the exponential function, is the working stability rate of the nozzle number at time t, with the superscript It is the mark of the number of times the variable is modified, and the subscript t represents the current processing time step or period sequence index.

[0010] As a further solution of the present invention, the nozzle number time difference is intended to filter out the nozzle number field, start time field and stop time field in the data, sort the data as a whole in the order of recording time, group the data according to the nozzle number field, read the corresponding start time and stop time values ​​one by one in the number group, perform numerical difference calculation to obtain the time difference of a single run, attach the corresponding number to the difference and mark it, repeat the traversal operation until the number processing is completed, classify and organize the time difference values ​​under all nozzle numbers according to the number dimension, and generate a time difference sequence set corresponding to the nozzle number.

[0011] As a further solution of the present invention, the specific steps of generating the delayed start delay adjustment scheme are: Based on the delay optimization model, the workload values ​​corresponding to the sprinkler numbers are extracted and a sprinkler and load comparison table is established. The delayed start time points corresponding to the load values ​​are matched. The start time sequence is sorted by number and the start interval boundaries are marked according to the load layer to generate a sprinkler start timing table. Based on the nozzle start timing table, extract the three data items of workload, injection pressure and inkjet volume from the corresponding numbered injection records, obtain the ratio of the injection pressure to the inkjet volume values ​​and generate a ratio sequence, filter out the data segments where the ratios undergo discontinuous jumps and extract the time index, locate the associated numbers to obtain the abnormal nozzle number set, and generate a nozzle load abnormality identification group; Based on the nozzle load abnormality identification group, the startup data and pressure data of the time period corresponding to the abnormal number are eliminated, and the startup time point and injection pressure range values ​​in the remaining numbers are retained. The startup time and the corresponding pressure value are extracted according to the number and the interval comparison and integration are performed. The startup delay and pressure range data group corresponding to the number is output to obtain the nozzle startup delay adjustment plan.

[0012] As a further solution of the present invention, the specific steps of generating the nozzle timing correction signal are: Based on the inkjet image captured in real time by the start-up delay adjustment scheme, a convolutional neural network is used to scan the row and column matrix of pixel values ​​in the image and extract boundary points with sudden grayscale changes, determine the connectivity of the boundary points and delineate the closed boundary contour, fit the angle of the boundary line segment, calculate the angle between the line segment and the line connecting the image center, and obtain the jet offset angle dataset; Based on the jetting deviation angle dataset, the nozzle number and angle difference are obtained, the vector modulus and direction angle of the angle difference are calculated, the trajectory center coordinate position is calibrated according to the pixel density distribution of the inkjet trajectory, and the Euclidean distance between the set target points is calculated. A parameter group consisting of the direction angle and the center offset is constructed to generate a nozzle deviation parameter set; Based on the nozzle deviation parameter set, the trigger point moment value and pulse width value in the current drive control signal of the nozzle are obtained, the increase or decrease of the trigger point value is adjusted according to the positive or negative direction angle, and the pulse width time interval is modified synchronously. After the nozzle adjustment, the data is numbered in sequence to obtain the nozzle timing correction signal.

[0013] As a further solution of the present invention, the convolutional neural network is based on the formula: ; in: is the injection deviation angle, is the inverse cosine function, is the direction vector of the boundary segment, is the direction vector from the image center to the midpoint of the boundary segment, is a vector With vector The dot product of is a vector The module length, is a vector The module length, is the first perturbation factor, is the contrast weight coefficient, is the grayscale mean difference between the areas on both sides of the boundary, is the Euclidean length of the boundary segment, is the second perturbation factor, is the gradient enhancement weight coefficient.

[0014] As a further solution of the present invention, the specific steps of generating the nozzle operation optimization model are: Based on the nozzle timing correction signal, the nozzle number and trigger adjustment value in the correction record are extracted, the execution time point of the print task corresponding to the number is retrieved and the response time before and after the correction is compared, the time sequence of the adjusted inkjet landing point is recorded, and the correspondence between the number and the landing point time is established to generate a nozzle operation reference map; Based on the nozzle operation reference map, the nozzle number, task start time and jet trigger time sequence in the printing task are extracted, the nozzle number sequence is arranged and the trigger time difference between adjacent numbers is calculated, a start interval matrix is ​​constructed and the valley time difference combination is extracted, the corresponding numbers of the combinations are sorted and a matching relationship index is established to obtain the start rhythm matching index; Based on the startup rhythm matching index, the historical inkjet records under the matching number are extracted and the time period with deviation marks is filtered, the correction trigger value and offset direction of the corresponding time are read, the correction records are classified and sorted by number and the offset amplitude data are summarized, and then the number matrix is ​​reorganized according to the rhythm order to obtain the nozzle operation optimization model.

[0015] A time delay system for nozzle printing software, the time delay system for nozzle printing software is used to execute the above-mentioned time delay method for nozzle printing software, the system comprising: Data sorting module: Based on the real-time working data of the sprinkler, the sprinkler number, start time and stop time in the record are extracted and sorted in ascending order of the time field. After grouping, the difference between the start and stop time is calculated by the sprinkler number. The start sequence and delay sequence are generated and marked according to the task period to generate the sprinkler operation time series; Timing modeling module: Based on the nozzle operation time series, the startup time difference data of the nozzle is extracted and the workload index is synchronously called to construct a calculation vector with the delay parameter as the core. After the difference offset is normalized to generate a period deviation vector set, the load normalization ratio of the aggregation number is formed into a delay evaluation matrix to generate the nozzle delay optimization model; State determination module: Based on the printhead delay optimization model, the module reads the delay trend and starting time point of the printhead number under multiple rounds of tasks, screens the floating data of the corresponding cycle pressure and inkjet volume, matches the different load stages by number, eliminates the cycle points with excessive pressure and retains the stable interval data, and generates a printhead start delay adjustment plan; Image correction module: Based on the nozzle start delay adjustment scheme, it synchronously executes the image acquisition command and obtains the inkjet image in the current cycle, performs edge segmentation and contour fitting on the ink droplet shape and position of the nozzle in the image, records the offset pixels between the actual nozzle injection position and the target image, constructs an offset comparison table, and generates a nozzle timing correction signal; Behavior optimization module: Based on the nozzle timing correction signal, the operation trajectory index is established according to the nozzle number, the corrected start and stop nodes in the continuous cycle are extracted, the historical ink deviation data in the print task log is used to calculate the nozzle spraying stability, and an operation habit template that associates the number and cycle is formed to generate a nozzle operation optimization model.

[0016] A time delay system for nozzle printing software, wherein the time delay system for nozzle printing software is used to execute the above-mentioned time delay method for nozzle printing software, and the system includes.

[0017] Compared with the prior art, the advantages and positive effects of the present invention are: 1. In the present invention, by numbering and classifying real-time data and splitting the start-stop behaviors item by item, it is possible to attribute the nozzle delay changes to stages at the micro level and form a structured time series under a unified dimension, laying the foundation for subsequent calculation accuracy.

[0018] 2. In the present invention, by analyzing the response behavior under different load environments, combining the inkjet volume and pressure difference to build a linkage judgment, delineating the abnormal interval and dynamically screening out the skewed data, the stability of the start-up delay adjustment is guaranteed, and the image error is used as the external observation channel to extract the inkjet deviation direction and pixel gap, reversely correct the nozzle operation time point, and establish a closed-loop control path at the time level.

[0019] 3. In the present invention, by coupling the three paths of real-time trajectory processing, difference matrix generation, image error mapping, and period correction, delay control no longer relies on fixed parameters or mean adjustment, and has dynamic mapping and fault isolation capabilities, greatly enhancing the flexibility, feedback granularity, and precision control of delay management. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 It is a schematic diagram of the workflow of the present invention; Figure 2 It is a system flow chart of the present invention. DETAILED DESCRIPTION

[0021] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0022] In the description of the present invention, it should be understood that the terms "length," "width," "up," "down," "front," "back," "left," "right," "vertical," "horizontal," "top," "bottom," "inside," "outside," and the like, indicating positions or relationships, are based on the positions or relationships shown in the accompanying drawings and are intended only to facilitate the description of the present invention and simplify the description. They do not indicate or imply that the devices or elements referred to must have a specific orientation, be constructed, or operate in a specific orientation. Therefore, they should not be construed as limiting the present invention. Furthermore, in the description of the present invention, "plurality" means two or more, unless otherwise expressly and specifically defined.

[0023] Example 1, please refer to Figure 1 The present invention provides a technical solution: a time delay method for nozzle printing software, comprising the following steps: Step 1: Sort the operation history data based on the real-time working data of the sprinkler, identify the delay changes during the operation of the sprinkler, extract the sprinkler timing characteristics and working status characteristics, generate the real-time operation time series of the sprinkler, and construct the sprinkler time series data set based on the time series data; Step 2: Based on the nozzle timing data set, the time difference between the start and stop of the nozzle is calculated and the delay optimization objective function is calculated. According to the model, the delay adjustment of the nozzle is calculated to generate the delay optimization model of the nozzle under different working conditions; Step 3: Based on the delay optimization model, obtain the start-up timing of the printhead under different workloads and analyze the relationship between printhead pressure and ink ejection volume to determine whether the printhead is overloaded or abnormal. Obtain the timing when the pressure is within the normal range and generate a start-up delay adjustment plan for the printhead; Step 4: Based on the inkjet image captured in real time by the start-up delay adjustment scheme, analyze the morphology and ejection direction of the inkjet nozzle, compare the difference between the actual inkjet effect and the expected image, and calculate the image error. Use the image recognition result to fine-tune the operation timing of the nozzle and generate a nozzle timing correction signal; Step 5: Track and optimize based on the nozzle timing correction signal, adjust the nozzle timing in combination with the inkjet feedback data after the printing task, obtain the nozzle operation reference map, start rhythm matching index and historical injection calibration matrix, and generate the nozzle operation optimization model.

[0024] The specific steps to generate the nozzle timing data set are: Based on the real-time working data of the sprinkler, the trigger instruction time field and the response feedback time field in the data are intercepted and the time difference is calculated. The sprinkler number field is used to classify each data and number them in time sequence. After arranging them in ascending order by the time field, a mapping table of sprinkler numbers and time differences is established to generate a sprinkler delay interval sequence; Based on the nozzle delay interval sequence, the delay value and the corresponding recorded injection frequency value of the nozzle are extracted and paired. The nozzle temperature data of the temperature change amplitude in the same time period is recorded. The injection event count value is collected and the event density per unit time is calculated. The three indicators are constructed into continuous vector sequences to establish the nozzle state feature matrix. Based on the nozzle state feature matrix, the time tags of all nozzle control instructions in the operation sequence are read, the control pulse width and trigger state mark at the time point are collected, and the injection confirmation flag related to the state change value is extracted. The data is aligned and summarized according to the time sequence to obtain the nozzle timing data set; Based on the real-time working data of the sprinkler, a sliding time window operation method is used, with a window length of 3 seconds and a step size of 0.5 seconds. The timestamp values ​​of the trigger instruction time field and the response feedback time field in the data are extracted respectively. The fields are processed using the time difference operation module, and the operation is T_diff_i=T_feedback_i-T_trigger_i, where T_trigger_i represents the trigger instruction time of the i-th record and T_feedback_i represents the response feedback time. The calculation result retains millisecond-level accuracy. All records are further classified and grouped by the nozzle_id field. Groupby(nozzle_id) is used to construct a number structured sequence. Then, rank(method='first') is executed on each grouped data to establish a time series number index. After sorting in ascending order by the time field T_trigger_i, zip(nozzle_id, T_diff_i) is called to construct a mapping structure between the sprinkler number and the time difference, generating a sprinkler delay interval sequence. Based on the nozzle delay interval sequence, a data pairing and merging method is used. Each delay record is bound to the firing_frequency field value under the corresponding nozzle number and a two-tuple V_pair = [T_diff_i, F_i] is constructed. F_i represents the firing frequency value of nozzle i in the corresponding cycle. The temperature field within the same cycle is recorded. The time alignment window is set to the maximum temperature difference within 0.2 seconds. Max(T_i) - min(T_i) is used to represent the temperature change amplitude ΔT_i. The firing event count field event_count is accumulated and calculated. The event density calculation formula D_i = count_i / 1s is executed within a fixed time slice (1 second). [T_diff_i, F_i], ΔT_i, and D_i are constructed as three independent vector sequences. They are merged into a matrix with a structural dimension of [N, 3] using np.stack(axis=1) to generate the nozzle status feature matrix. Based on the nozzle status feature matrix, the label sequence alignment method is adopted to extract the command_timestamp field of all nozzle operation records in the control log, construct the control instruction time label sequence T_control_seq, synchronously read the pulse_width field of each record as the control pulse width scalar W_i, collect the trigger status mark field status_flag_i as the operation status type label, read the confirmation bit field confirm_bit_i for each ejection record to indicate whether the ink droplet ejection is actually triggered, and perform a unified time axis alignment operation on the four data according to T_control_seq. The merge_asof(on=T_control_seq) method is used to fuse all feature columns to obtain a time-aligned joint operation sequence, and the numbering dimension is grouped by the numbering field to generate a nozzle timing dataset.

[0025] The specific steps to generate a delay optimization model are: Based on the sprinkler timing data set, the numerical difference between the start time field and the stop time field in the record is calculated, the time difference values ​​of the sprinkler numbers are classified and sorted, and the data are arranged in order using single numbers and the extreme values ​​and arithmetic mean values ​​are calculated to generate a sprinkler operation cycle statistics table; Based on the statistics table of the nozzle operation cycle, the difference in average operation time is calculated and a number index table is established. The period item corresponding to the peak deviation number is extracted and the absolute value of the deviation is calculated. The interval range is established by combining the valley period and the peak period to form a difference matrix, and the nozzle delay difference vector set is established. Based on the nozzle delay difference vector set, the workload value under the nozzle number is extracted using the long short-term memory network and normalized ratio conversion is performed on the cycle deviation value. The ratio set of the same number under different temperature conditions and trigger frequencies is sorted, and the number is divided into state categories and the corresponding value of each state is filled in to obtain the nozzle delay optimization model. Based on the nozzle time series dataset, the field difference operation method is used to extract the start_time and stop_time field values ​​in each record. The custom function diff_time=stop_time-start_time is called to calculate the time difference. The time format is set to UNIX millisecond timestamp, and the return value is rounded to three decimal places. The nozzle_id field is numbered and classified, and the number is used as the primary key for grouping. The time difference sequence in each group is sorted in ascending order on the time axis using sort_values(by=start_time, ascending=True). Then, the three aggregation functions min(), max(), and mean() are used to calculate the minimum, maximum, and arithmetic mean of each group respectively. The three value results are constructed into a field column, and the number column is spliced ​​as the primary key index to generate a nozzle operation cycle statistics table. Based on the nozzle operation cycle statistics table, the extreme value deviation calculation method is used, the mean_cycle_time field is called as the reference value of the average operation time, the range is calculated by the field max_cycle_time-min_cycle_time and used as the cycle floating range, and the fluctuation screening threshold is set to the average time difference of ±10%. The logical screening expression abs(cycle_time-mean_cycle_time)>0.1 Extract the out-of-limit sprinkler number using mean_cycle_time, use the cycle position corresponding to the abnormal number as the index to establish a number index table, identify the maximum term peak_index and the minimum term valley_index of the deviation cycle, calculate the absolute difference as abs(peak_cycle-valley_cycle), and then use range(peak_index, valley_index+1) to construct the cycle interval. Extract the time difference of each cycle in the interval in turn and store them in a two-dimensional matrix structure. The matrix index is the sprinkler number and the column dimension is the cycle sequence, thus generating a sprinkler delay difference vector set. Based on the nozzle delay difference vector set, the long short-term memory network model structure is constructed using the long short-term memory network method. The input dimension is set to [batch_size, time_steps, feature_dim], where feature_dim contains four parameters: the delay vector corresponding to the nozzle number, the cycle deviation, the trigger frequency, and the ambient temperature. The vector set is constructed into sample batch data using a sliding window method. The window step size is set to 1 and the window width is set to 8 cycles. The input data is converted into a three-dimensional tensor format [N, 8, 4] and input into the model. The predicted workload value L_pred_i corresponding to each time step is output, and then a normalized ratio conversion is performed. The formula R_i=L_pred_i / P_i is used, where P_i represents the cycle deviation value. The results are taken from the original periodic table and aligned by number. They are classified according to the number dimension. The normalized results are recorded in the two-dimensional mapping table constructed under the temperature interval and frequency interval respectively. The ratios are grouped by the status label field status_type to fill the state matrix to generate the nozzle delay optimization model.

[0026] Long short-term memory network, according to the formula: ; in: is the normalized ratio after nozzle delay optimization, is the attention weight coefficient of the numbered nozzle at time t, is the workload prediction value output by the long short-term memory network at time t for the sprinkler number, is the load correction weight coefficient of the ambient temperature parameter T, where T is the working ambient temperature of the current nozzle number. is the load correction weight coefficient of the trigger frequency parameter f, f is the current nozzle trigger frequency, is the delay fluctuation parameter The load correction weight coefficient, is the maximum delay time difference of the nozzle in the past three cycles, is the cooling rate parameter The load correction weight coefficient, is the cooling rate of the nozzle temperature drop, is the load standard deviation The penalty adjustment coefficient, is the workload standard deviation of the sprinkler number at time t, is the periodic deviation value of the nozzle number at time t, is the period deviation zero value compensation constant, is the modulation coefficient of the exponential deviation correction term, is the natural exponential function, is the attenuation coefficient in the exponential function, is the working stability rate of the nozzle number at time t, with the superscript It is the mark of the number of times the variable is modified, and the subscript t represents the current processing time step or period sequence index; Execution process: Collect the control input sequence of the nozzle under multiple consecutive printing cycles, input it into the long short-term memory network model to extract the workload prediction value of the corresponding time step, use the state classifier to generate the attention weight of the current nozzle number in the cycle, which is used to measure the task priority of the nozzle in the current state, obtain the current ambient temperature, trigger frequency, delay fluctuation value and cooling rate of the nozzle, introduce the corresponding weight coefficients in turn, construct the load dynamic adjustment factor, correct the load value expression range, extract the load standard deviation in the current cycle of the numbered nozzle, combine the coefficients to form a stability penalty term, and superimpose it on the overall load adjustment expression. At the same time, the cycle deviation value is used as the basic value of the denominator, introduce the offset correction amount and the exponential form of the stability rate penalty correction, complete the nonlinear slow release modeling of the periodic disturbance, integrate all terms into the normalized ratio function, complete the nozzle delay ratio calculation under multi-state input conditions, and generate high-dimensional available delay optimization features for the nozzle control system.

[0027] The time difference of the nozzle number is designed to filter out the nozzle number field, start time field and stop time field in the data, sort the data as a whole in the order of recording time, group the data by the nozzle number field, read the corresponding start time and stop time values ​​one by one in the number group, calculate the numerical difference to obtain the time difference of a single run, add the corresponding number to the difference and mark it, repeat the traversal operation until the number processing is completed, classify and organize the time difference values ​​under all nozzle numbers according to the number dimension, and generate a time difference sequence set corresponding to the nozzle number.

[0028] The specific steps to generate a delayed start delay adjustment plan are: Based on the delay optimization model, the workload values ​​corresponding to the sprinkler numbers are extracted and a sprinkler-load comparison table is established. The delayed start time points corresponding to the load values ​​are matched. The start time sequence is sorted by number and the start interval boundaries are marked according to the load layer to generate a sprinkler start timing table. Based on the nozzle start timing table, the workload, injection pressure, and inkjet volume data of the corresponding injection records are extracted. The ratio of the injection pressure to the inkjet volume is obtained and a ratio sequence is generated. The data segments where the ratio has discontinuous jumps are screened out and the time index is extracted. The associated numbers are located to obtain the abnormal nozzle number set and generate the nozzle load abnormality identification group; Based on the nozzle load abnormality identification group, the startup data and pressure data of the time period corresponding to the abnormal number are eliminated, and the startup time point and injection pressure range values ​​of the remaining numbers are retained. The startup time and the corresponding pressure value are extracted by number and integrated by interval comparison. The startup delay and pressure range data group corresponding to the number is output to obtain the nozzle startup delay adjustment plan; Based on the delay optimization model, the number vector mapping method is used to extract the field nozzle_id and the corresponding field workload_value. zip(nozzle_id, workload_value) is used to construct a number and workload mapping table. Groupby(nozzle_id) is called to classify and group the workload within the number. Merge_asof() is used to connect the field startup_delay_point by the timestamp field, and the load value of each number is mapped to the delayed startup point. Sort_values(by=startup_delay_point) is called to sort the startup time series within the number in ascending order. The load stratification standard is set to a fluctuation of 10% above and below the mean. The logical expression layer_1=workload_value<=avg-0.1 is used. avg,layer_2=workload_value>=avg+0.1 avg establishes the interval boundary, calls pd.cut() to divide the start time points within the number into layered segments, and generates a nozzle start timing table; Based on the nozzle start timing table, a jump detection method is used to extract the fields workload_value, spray_pressure, and ink_volume from the nozzle number mapping. The custom function ratio=spray_pressure / ink_volume is called to calculate the ratio sequence for each record. The continuous ratio sequence is constructed as a vector R_i. The jump judgment window width is set to 3, and the detection rule is abs(R_i-R_{i-1})>0.2. R_{i-1}, perform Boolean mask filtering on abnormal points, record the timestamp index jump_index of the jump position, and extract the corresponding nozzle_id number combination as an abnormal number set. After deduplication, output the number field set to generate the nozzle load abnormal identification group; Based on the nozzle load anomaly identification group, the anomaly number exclusion method is used to remove the anomaly number and its corresponding time period data from the startup_time and spray_pressure sequences. The remaining data is numbered and classified using groupby(nozzle_id). The time fields startup_time and spray_pressure_range are extracted from each group of data for one-to-one mapping. The zip() function is called to merge the two items into a control group structure. A mapping dictionary structure is established by number, and all numbered time points and pressure ranges are integrated into the output values ​​of the startup_delay_adjustment and pressure_scope fields to generate a nozzle startup delay adjustment plan.

[0029] The specific steps to generate the nozzle timing correction signal are as follows: Based on the inkjet image captured in real time with the start-up delay adjustment scheme, a convolutional neural network is used to scan the row and column matrix of pixel values ​​in the image and extract the boundary points with sudden grayscale changes. The connectivity of the boundary points is determined and the closed boundary contour is delineated. The angle of the boundary segment is fitted and the angle between the segment and the line connecting the image center is calculated to obtain the jet offset angle dataset. Based on the jetting deviation angle dataset, the nozzle number and angle difference are obtained, the vector modulus and direction angle of the angle difference are calculated, the trajectory center coordinate position is calibrated according to the pixel density distribution of the inkjet trajectory, and the Euclidean distance between the set target points is calculated. A parameter group consisting of the direction angle and the center offset is constructed to generate a nozzle deviation parameter set; Based on the nozzle deviation parameter set, the trigger point moment value and pulse width value in the current drive control signal of the nozzle are obtained. The trigger point value increase or decrease is adjusted according to the positive or negative direction angle and the pulse width time interval is modified synchronously. The nozzle adjustment data is integrated and corrected according to the sequential numbering to obtain the nozzle timing correction signal; Based on the inkjet image captured in real time by the start-up delay adjustment scheme, a convolutional neural network method is used to scan the pixel matrix using a three-layer convolutional network model. The first layer uses Conv2D (filters=32, kernel_size=3, activation='relu') to extract the first-order changes in the image grayscale value. The second layer uses Conv2D (filters=64, kernel_size=3, activation='relu') to detect edge contour change points, and uses MaxPooling2D (pool_size=2) to perform The row region is compressed, and the output feature map tensor size is [batch_size, height / 2, width / 2, 64]. The pixel grayscale gradient comparison operation is performed on the feature map. The pixel pairs with grayscale difference exceeding the set threshold ΔI>25 are identified as boundary points. The connectedComponents() function is called to analyze the connectivity of the boundary pixels. The closed contour areas in the connected areas are screened out. The least squares method is used to fit five consecutive pixels in the boundary contour. After fitting the line segment, the angle between the line segment and the line connecting the geometric center of the image is calculated. The number and angle pair set are output to generate the injection offset angle dataset. Based on the jet offset angle dataset, vector operation and trajectory analysis methods are used to extract the angle field θ_i from the image feature points and the nozzle number nozzle_id to form a number vector pair. The vector modulus of the angle difference is calculated by calling the formula Mi_i=sqrt((Δx)^2+(Δy)^2). The direction angle uses the formula θ_i=arctan2(Δy, Δx) to represent the inkjet direction offset. The density distribution of the ink pixel position in the image is calculated separately, and the pixels per unit square are set to 10×10 , count the number of pixels in each block as the local density, and take the weighted average of the local density to obtain the center of gravity coordinates (x_c, y_c) of the inkjet trajectory. Then, with the nozzle set target point coordinates (x_0, y_0), the Euclidean distance formula D_i=sqrt((x_c-x_0)^2+(y_c-y_0)^2) is used to calculate the offset. Finally, the number field, direction angle and center of gravity offset value are combined to output the structure [nozzle_id, θ_i, D_i] to generate the nozzle deviation parameter set; Based on the nozzle deviation parameter set, the timing signal synchronization adjustment method is adopted. The fields trigger_time_i and pulse_width_i in the nozzle control data table are read, and a number matching index is established according to nozzle_id. The positive and negative values ​​of θ_i are used as the increase or decrease signal adjustment flag. If θ_i>0, trigger_time_i=trigger_time_i-δt is executed, where δt=round(abs(θ_i)×0.6) is the time adjustment amount. If θ_i<0, trigger_time_i=trigger_time_i+δt is executed. At the same time, the output pulse interval of pulse_width_i is adjusted according to the offset distance D_i. The formula is set to pulse_width_i=pulse_width_i+α×D_i, where α=0.25, which is used to scale the time window according to the unit pixel offset ratio. Finally, the three columns of nozzle_id, trigger_time_i and pulse_width_i are merged and sorted to output the number timing table to generate the nozzle timing correction signal.

[0030] Convolutional neural network, according to the formula: ; in: is the injection deviation angle, is the inverse cosine function, is the direction vector of the boundary segment, is the direction vector from the image center to the midpoint of the boundary segment, is a vector With vector The dot product of is a vector The module length, is a vector The module length, is the first perturbation factor, is the contrast weight coefficient, is the grayscale mean difference between the areas on both sides of the boundary, is the Euclidean length of the boundary segment, is the second perturbation factor, is the gradient enhancement weight coefficient.

[0031] Execution process: Call the convolutional neural network to perform pixel matrix scanning on the inkjet image captured in real time, extract the grayscale mutation boundary points and complete the closed contour construction, approximate the closed boundary contour into multiple line segments, extract the direction vector for each boundary segment, and calculate the direction vector formed by the midpoint of the segment to the image center, calculate the dot product between and, respectively obtain and, and introduce the disturbance term to correct the denominator to avoid zero value anomalies. In order to improve the accuracy of image boundary recognition, further extract the grayscale mean difference and segment length of the area on both sides of the boundary segment, superimpose the disturbance term to form a gradient factor, multiply it by the weight coefficient and gradient response weight coefficient determined by contrast normalization analysis, respectively, combine the two parts of the result and input them into the inverse cosine function to calculate the improved jet offset angle, and finally construct the data set required for nozzle delay calibration and use it for inkjet trajectory control.

[0032] The specific steps to generate the nozzle operation optimization model are: Based on the nozzle timing correction signal, the nozzle number and trigger adjustment value in the correction record are extracted, the execution time point of the number corresponding to the printing task is retrieved and the response time before and after the correction is compared. The adjusted inkjet landing time sequence is recorded and the corresponding relationship between the number and the landing time is established to generate a nozzle operation reference map; Based on the nozzle operation reference map, the nozzle number, task start time and jet trigger time sequence in the printing task are extracted. The nozzle numbers are arranged in order and the trigger time difference between adjacent numbers is calculated. The start interval matrix is ​​constructed and the valley time difference combination is extracted. The corresponding numbers of the combinations are sorted and a matching relationship index is established to obtain the start rhythm matching index. Based on the startup rhythm matching index, the historical inkjet records under the matching number are extracted and the time period with deviation marks is screened. The correction trigger value and offset direction of the corresponding time are read. The correction records are classified and sorted by number and the offset amplitude data is summarized. The number matrix is ​​then reorganized according to the rhythm sequence to obtain the nozzle operation optimization model. Based on the nozzle timing correction signal, using the landing point sequence mapping method, extract the fields nozzle_id and adjustment_value, execute merge(nozzle_id, adjustment_value) to construct a comparison table of nozzle numbers and trigger adjustment values, call the timestamp in the field task_execution_time and perform a time difference operation on the two columns of data response_time_before and response_time_after, call adjusted_point_time = response_time_after - adjustment_value, retain the adjusted inkjet time point, establish a mapping structure dict(zip(nozzle_id, adjusted_point_time)), construct a mapping table of numbers and landing points and sort the results in ascending order of numbers, and generate a nozzle operation reference map; Based on the nozzle operation reference map, using the difference matrix construction method, read the fields nozzle_id, task_start_time, trigger_time, perform a sorting order on the numbers by sort_values(by = nozzle_id), call np.diff(trigger_time_array) to perform a difference calculation on the trigger times between adjacent numbers, construct a two-dimensional array interval_matrix[i][j]=trigger_time[j]-trigger_time[i], set the threshold condition interval_diff < quantile(0.1) as the valley value criterion, filter the number pairs that meet the conditions and construct dict(zip(nozzle_pair, interval_diff)), sequentially extract the number sequence and the interval value sequence to establish a primary key index structure, and generate a startup rhythm matching index; Based on the startup rhythm matching index, the rhythm sequence alignment method is used to extract the fields ink_log_time and offset_flag under the matching number, set the mask condition offset_flag==1 to filter the deviation records, extract its time point field offset_time and direction_value field to establish a combined vector offset_set=[offset_time, direction_value], call groupby(nozzle_id) to classify and organize the offset direction data, execute sum(abs(direction_value)) to calculate the offset amplitude value within the number, set the rhythm sequence structure to sorted(interval_index.keys()), and reorganize it into a matrix structure adjusted_matrix[i][j]=offset_set[j] according to the number arrangement to generate the nozzle operation optimization model.

[0033] See also Figure 2 , a time delay system for nozzle printing software, the system includes: Data sorting module: Based on the real-time working data of the sprinkler, the sprinkler number, start time and stop time in the record are extracted and sorted in ascending order of the time field. After grouping, the difference between the start and stop time is calculated by the sprinkler number. The start sequence and delay sequence are generated and marked according to the task period to generate the sprinkler operation time series; Timing modeling module: Based on the nozzle operation time series, the startup time difference data of the nozzle is extracted and the workload index is synchronously called. A calculation vector with delay parameters as the core is constructed. After the difference offset is normalized to generate a period deviation vector set, the load normalization ratio of the aggregation number is formed into a delay evaluation matrix to generate the nozzle delay optimization model. State determination module: Based on the printhead delay optimization model, it reads the delay trend and starting time point of the printhead number under multiple rounds of tasks, screens the floating data of the corresponding cycle pressure and inkjet volume, matches different load stages by number, eliminates the cycle points with excessive pressure and retains the stable interval data, and generates the printhead start delay adjustment plan; Image correction module: Based on the nozzle start delay adjustment scheme, it synchronously executes the image acquisition command and obtains the inkjet image in the current cycle. It performs edge segmentation and contour fitting on the ink droplet shape and position of the nozzle in the image, records the offset pixels between the actual nozzle injection position and the target image, constructs an offset comparison table, and generates a nozzle timing correction signal. Behavior optimization module: Based on the nozzle timing correction signal, the operation trajectory index is established according to the nozzle number, the corrected start and stop nodes in the continuous cycle are extracted, and the historical ink deviation data in the print task log is used to calculate the nozzle injection stability. An operation habit template that associates the number and cycle is formed to generate a nozzle operation optimization model.

[0034] The above are merely preferred embodiments of the present invention and do not limit the present invention in any other form. Any technician familiar with the profession may use the technical content disclosed above to change or modify it into an equivalent embodiment with equivalent changes and apply it to other fields. However, any simple modification, equivalent change and modification made to the above embodiment based on the technical essence of the present invention without departing from the content of the technical solution of the present invention shall still fall within the scope of protection of the technical solution of the present invention.

Claims

1. A time delay method for nozzle printing software, characterized in that: The following steps are involved: Step 1: Sort the operation history data based on the real-time working data of the sprinkler, identify the delay changes during the operation of the sprinkler, extract the sprinkler timing characteristics and working status characteristics, generate the real-time operation time series of the sprinkler, and construct the sprinkler time series data set based on the time series data; Step 2: Based on the nozzle timing data set, calculate the time difference between the start and stop of the nozzle and calculate the delay optimization objective function, calculate the delay adjustment of the nozzle according to the model, and generate the delay optimization model of the nozzle under different working conditions; Step 3: Based on the delay optimization model, the start-up timing of the printhead under different workloads is obtained and the relationship between the printhead pressure and the ink ejection volume is analyzed to determine whether the printhead is overloaded or abnormal. The timing when the pressure is within the normal range is obtained, and a start-up delay adjustment plan for the printhead is generated; Step 4: Based on the inkjet image captured in real time by the start-up delay adjustment scheme, analyze the morphology and ejection direction of the inkjet printhead, compare the difference between the actual inkjet effect and the expected image, calculate the image error, use the image recognition result to fine-tune the operation timing of the printhead, and generate a printhead timing correction signal; Step 5: Track and optimize based on the nozzle timing correction signal, adjust the nozzle timing in combination with the inkjet feedback data after the printing task, obtain the nozzle operation reference map, start rhythm matching index and historical injection calibration matrix, and generate a nozzle operation optimization model.

2. The time delay method for nozzle printing software according to claim 1, characterized in that: The specific steps for generating the nozzle timing data set are: Based on the real-time working data of the sprinkler, the trigger instruction time field and the response feedback time field in the data are intercepted and the time difference is calculated. The sprinkler number field is used to classify each data and number them in time sequence. After arranging them in ascending order by the time field, a mapping table of sprinkler numbers and time differences is established to generate a sprinkler delay interval sequence; Based on the nozzle delay interval sequence, the delay value and the corresponding recorded injection frequency value of the nozzle are extracted and paired, the nozzle temperature data of the temperature change amplitude in the same time period is recorded, the injection event count value is collected and the event density per unit time is calculated, and the three indicators are constructed into a continuous vector sequence to establish the nozzle state feature matrix; Based on the nozzle state feature matrix, the time tags of all nozzle control instructions in the operation sequence are read, the control pulse width and trigger state mark at the time point are collected, the injection confirmation flag related to the state change value is extracted, and the data is aligned and summarized according to the time sequence to obtain the nozzle timing data set.

3. The time delay method for nozzle printing software according to claim 1, characterized in that: The specific steps of generating the delay optimization model are: Based on the nozzle timing data set, the numerical difference between the start time field and the stop time field in the record is calculated, the time difference values ​​of the nozzle numbers are classified and sorted, and the data are arranged in order using single numbers and the extreme values ​​and arithmetic mean values ​​of the data are calculated to generate a nozzle operation cycle statistics table; Based on the nozzle operation cycle statistics table, the difference in average operation time is calculated and a number index table is established. The period item corresponding to the peak deviation number is extracted and the absolute value of the deviation is calculated. The interval range is established by combining the valley period and the peak period and forming a difference matrix to establish a nozzle delay difference vector set; Based on the nozzle delay difference vector set, the long short-term memory network is used to extract the workload value under the nozzle number and perform normalized ratio conversion with the period deviation value. The ratio sets of the same number under different temperature conditions and trigger frequencies are sorted out, and the numbers are divided into state categories and filled with corresponding values ​​in each state to obtain the nozzle delay optimization model.

4. The time delay method for nozzle printing software according to claim 3, characterized in that: The long short-term memory network is based on the formula: Where: η″ is the normalized ratio after nozzle delay optimization, α t is the attention weight coefficient of the numbered nozzle at time t, h t is the workload prediction value output by the long short-term memory network at the time t for the nozzle number, λ1 is the load correction weight coefficient of the ambient temperature parameter T, T is the working environment temperature of the current nozzle number, λ2 is the load correction weight coefficient of the trigger frequency parameter f, f is the current nozzle trigger frequency, λ3 is the load correction weight coefficient of the delay fluctuation parameter Δd, Δd is the maximum delay time difference of the nozzle in the past three cycles, λ4 is the load correction weight coefficient of the cooling rate parameter ρ, ρ is the cooling rate of the nozzle temperature drop, and μ is the load standard deviation σ t The penalty adjustment coefficient, σ t is the workload standard deviation of the sprinkler number at time t, δ t is the periodic deviation value of the nozzle number at time t, ∈ is the periodic deviation zero value compensation constant, ξ is the modulation coefficient of the exponential deviation correction term, exp is the natural exponential function, γ is the attenuation coefficient in the exponential function, R t is the working stability rate of the nozzle number at time t, the superscript " is the variable correction times mark, and the subscript t represents the current processing time step or cycle sequence index.

5. According to the delay method for nozzle printing software described in claim 3, the nozzle number time difference is intended to filter out the nozzle number field, start time field and stop time field in the data, sort the data as a whole in the order of recording time, group the data according to the nozzle number field, read the corresponding start time and stop time values ​​one by one in the number group, perform numerical difference calculation to obtain the single run time difference, add the corresponding number to the difference and mark it, repeat the traversal operation until the number processing is completed, classify and organize the time difference values ​​under all nozzle numbers according to the number dimension, and generate a time difference sequence set corresponding to the nozzle number.

6. The time delay method for nozzle printing software according to claim 1, characterized in that: The specific steps for generating the delayed start delay adjustment scheme are: Based on the delay optimization model, the workload values ​​corresponding to the sprinkler numbers are extracted and a sprinkler and load comparison table is established. The delayed start time points corresponding to the load values ​​are matched. The start time sequence is sorted by number and the start interval boundaries are marked according to the load layer to generate a sprinkler start timing table. Based on the nozzle start timing table, extract the three data items of workload, injection pressure and inkjet volume from the corresponding numbered injection records, obtain the ratio of the injection pressure to the inkjet volume values ​​and generate a ratio sequence, filter out the data segments where the ratios undergo discontinuous jumps and extract the time index, locate the associated numbers to obtain the abnormal nozzle number set, and generate a nozzle load abnormality identification group; Based on the nozzle load abnormality identification group, the startup data and pressure data of the time period corresponding to the abnormal number are eliminated, and the startup time point and injection pressure range values ​​in the remaining numbers are retained. The startup time and the corresponding pressure value are extracted according to the number and the interval comparison and integration are performed. The startup delay and pressure range data group corresponding to the number is output to obtain the nozzle startup delay adjustment plan.

7. The time delay method for nozzle printing software according to claim 1, characterized in that: The specific steps of generating the nozzle timing correction signal are as follows: Based on the inkjet image captured in real time by the start-up delay adjustment scheme, a convolutional neural network is used to scan the row and column matrix of pixel values ​​in the image and extract boundary points with sudden grayscale changes, determine the connectivity of the boundary points and delineate the closed boundary contour, fit the angle of the boundary line segment, calculate the angle between the line segment and the line connecting the image center, and obtain the jet offset angle dataset; Based on the jetting deviation angle dataset, the nozzle number and angle difference are obtained, the vector modulus and direction angle of the angle difference are calculated, the trajectory center coordinate position is calibrated according to the pixel density distribution of the inkjet trajectory, and the Euclidean distance between the set target points is calculated. A parameter group consisting of the direction angle and the center offset is constructed to generate a nozzle deviation parameter set; Based on the nozzle deviation parameter set, the trigger point moment value and pulse width value in the current drive control signal of the nozzle are obtained, the increase or decrease of the trigger point value is adjusted according to the positive or negative direction angle, and the pulse width time interval is modified synchronously. After the nozzle adjustment, the data is numbered in sequence to obtain the nozzle timing correction signal.

8. The time delay method for nozzle printing software according to claim 7, characterized in that: The convolutional neural network is based on the formula: Where: θ′ is the injection deviation angle, arccos is the inverse cosine function, is the direction vector of the boundary segment, is the direction vector from the image center to the midpoint of the boundary segment, is a vector With vector The dot product of is a vector The module length, is a vector The modulus of ∈1 is the first perturbation factor, w contrast is the contrast weight coefficient, ΔI is the grayscale mean difference between the two sides of the boundary, Δd is the Euclidean length of the boundary segment, ∈2 is the second perturbation factor, and w gradient is the gradient enhancement weight coefficient.

9. The time delay method for nozzle printing software according to claim 1, characterized in that: The specific steps of generating the nozzle operation optimization model are: Based on the nozzle timing correction signal, the nozzle number and trigger adjustment value in the correction record are extracted, the execution time point of the print task corresponding to the number is retrieved and the response time before and after the correction is compared, the time sequence of the adjusted inkjet landing point is recorded, and the correspondence between the number and the landing point time is established to generate a nozzle operation reference map; Based on the nozzle operation reference map, the nozzle number, task start time and jet trigger time sequence in the printing task are extracted, the nozzle number sequence is arranged and the trigger time difference between adjacent numbers is calculated, a start interval matrix is ​​constructed and the valley time difference combination is extracted, the corresponding numbers of the combinations are sorted and a matching relationship index is established to obtain the start rhythm matching index; Based on the startup rhythm matching index, the historical inkjet records under the matching number are extracted and the time period with deviation marks is filtered, the correction trigger value and offset direction of the corresponding time are read, the correction records are classified and sorted by number and the offset amplitude data are summarized, and then the number matrix is ​​reorganized according to the rhythm order to obtain the nozzle operation optimization model.

10. A time delay system for nozzle printing software, characterized in that: The time-delay method for nozzle printing software according to any one of claims 1 to 9, wherein the system comprises: Data sorting module: Based on the real-time working data of the sprinkler, the sprinkler number, start time and stop time in the record are extracted and sorted in ascending order of the time field. After grouping, the difference between the start and stop time is calculated by the sprinkler number. The start sequence and delay sequence are generated and marked according to the task period to generate the sprinkler operation time series; Timing modeling module: Based on the nozzle operation time series, the startup time difference data of the nozzle is extracted and the workload index is synchronously called to construct a calculation vector with the delay parameter as the core. After the difference offset is normalized to generate a period deviation vector set, the load normalization ratio of the aggregation number is formed into a delay evaluation matrix to generate the nozzle delay optimization model; State determination module: Based on the printhead delay optimization model, the module reads the delay trend and starting time point of the printhead number under multiple rounds of tasks, screens the floating data of the corresponding cycle pressure and inkjet volume, matches the different load stages by number, eliminates the cycle points with excessive pressure and retains the stable interval data, and generates a printhead start delay adjustment plan; Image correction module: Based on the nozzle start delay adjustment scheme, it synchronously executes the image acquisition command and obtains the inkjet image in the current cycle, performs edge segmentation and contour fitting on the ink droplet shape and position of the nozzle in the image, records the offset pixels between the actual nozzle injection position and the target image, constructs an offset comparison table, and generates a nozzle timing correction signal; Behavior optimization module: Based on the nozzle timing correction signal, the operation trajectory index is established according to the nozzle number, the corrected start and stop nodes in the continuous cycle are extracted, the historical ink deviation data in the print task log is used to calculate the nozzle spraying stability, and an operation habit template that associates the number and cycle is formed to generate a nozzle operation optimization model.

Images