Printers with a first feed roller, adjustment devices, and printer adjustment methods.

Abstract

In the present invention, when an individual graph data set is denoted by gr, an initial value thereof is denoted by gri, an average of graph data sets in the whole of a ductor roller is denoted by g, and an initial value thereof is denoted by gi, graph data sets gr and g are corrected during printing so as to eliminate errors between a printing density and a target density. When a stabilization value of the average graph data set g is denoted by ge, and a stabilization value of the individual graph data set gr is denoted by gre, data sets including the initial values gi, gri and the stabilization values ge, gre are collected. A base parameter B is increased or decreased on the basis of the difference, in the collected data sets, between the distribution of the stabilization value ge and the distribution of the initial value gi. The collected data sets are divided on the basis of each printing speed, and, with respect to a speed parameter V, a speed parameter V for each division is increased or decreased, on the basis of the difference, in each division of the graph data set g, between the distribution of the stabilization value ge and the distribution of the initial value gi. The collected data sets are divided on the basis of the average graph data set g, and an area parameter F for each g division is increased or decreased on the basis of the difference, in each g division, between the distribution of the stabilization value ge and the distribution of the initial value gi. A roller parameter R of a corresponding individual roller is increased or decreased on the basis of the difference, in the individual graph data set, between the distribution of the stabilization value gre and the distribution of the initial value gri. The duty ratio of the ductor roller is corrected on the basis of each parameter value.

Description

Printing press equipped with duct roller, correction device, and correction method for printing press This invention relates to a printing press equipped with a duct roller. In printing presses equipped with duct rollers, the duct rollers are positioned between the kettle roller and the mixing roller. The duct rollers are divided into multiple sections along the axial direction, and the duty cycle of the time each individual roller is in contact with the kettle roller can be controlled. The print density for each color in the printed material is then measured, and the individual rollers in the duct rollers are feedback-controlled to match the target density. This reduces fluctuations in print density during printing (Patent Document 1: JP2015-63071A, corresponding US9446581). The duct roller contains ink that has already been drawn from the kettle roller, and there is also ink present on the mixing roller. Because of this, there is a delay in controlling the print density using the duct roller. For this reason, it is known that when changing the plate and increasing the image area ratio, the amount of ink drawn to the duct roller is temporarily increased, and when decreasing the image area ratio, the amount of ink drawn to the duct roller is temporarily decreased (Patent Document 1: JP2015-63071A, corresponding US9446581). JP2015-63071A, corresponding to US9446581 By monitoring the print density and providing feedback to the duct roller, the control data for the duct roller is optimized. However, how to utilize this data for printing on subsequent days has not been considered. For example, if printing with the same plate the next day, the previous day's data could be used again. However, this is rare, and how to carry over the data obtained up to the previous day when printing with a different plate the next day has not been considered. Since data from the previous plate cannot be used, and each time the plate is changed, the amount of wasted paper increases if the printing conditions are monitored until the print density reaches an acceptable range. Also, inexperienced operators will produce more wasted paper, so the system becomes dependent on skilled operators. This invention corrects the control data of the duct roller by having the correction device learn, ・ Reduce losses such as waste paper at the start of printing, ・ Improve print quality, ・ Correct fluctuations in the state of the printing press, ・ Make it unnecessary to rely on more skilled operators, which is an issue. The printing press equipped with the doctor roller of this invention includes an ink fountain, a fountain roller that contacts the ink fountain, a doctor roller, a kneading roller, and a control device that controls the doctor roller, Along the axial direction of the doctor roller, a plurality of individual rollers are arranged, Taking the contact time τ between the individual roller and the fountain roller and the control period T1 of the individual roller, the control device is configured to control the ink extraction amount of the individual roller by controlling the duty ratio consisting of the ratio τ / T1 of the contact time to the control period, Let the target value of the ink supply amount at the individual roller (individual graph data) be gr, the initial value gri of gr determined from the image to be printed, the average value (average graph data) of gr of the entire doctor roller of the individual graph data gr be g, and the initial value gi of g. The control device controls the duty ratio of the individual roller based on the individual graph data gr, and is configured to change the individual graph data gr so as to eliminate the error between the actually measured print density and the target print density, or according to the operator's input, The printing press further includes a correction device that corrects the duty ratio. The correction device is Taking the stable value ge of the average graph data g and the stable value gre of the individual graph data gr, collecting data including the initial values gi, gri and the stable values ge, gre of the graph data g, gr, Updating the base parameter B based on the difference between the distribution of the stable value ge of the average graph data and the distribution of the initial value gi in the collected data, For the speed parameter V, which is a parameter for each section of the printing speed when dividing the collected data into each range of the printing speed, updating the speed parameter V based on the difference between the distribution of the stable value ge of the average graph data and the distribution of the initial value gi in each section of the printing speed. When dividing the collected data into ranges of average graph data g, the area parameter F, which is a parameter for each range of average graph data g, is updated for each range of average graph data g based on the difference between the distribution of the stable value ge of the average graph data and the distribution of the initial value gi. The collected data is processed for each individual roller, and the roller parameter R, which is a parameter for each individual roller, is updated based on the difference between the distribution of the stable value gre and the distribution of the initial value gri in the individual graph data. Based on the updated base parameter B, the speed parameter V corresponding to the printing speed, and the area parameter F corresponding to the average graph data g, the duty cycle of all individual rollers is changed. The system is configured to change the duty cycle of the corresponding individual roller based on the updated roller parameter R. In the correction device and correction method for a printing press of this invention, The system includes an ink fountain, a fountain roller that contacts the ink fountain, a duct roller, a mixing roller, and a control device that controls the duct roller. Multiple individual rollers are arranged along the axial direction of the duct roller. The control device is configured to control the amount of ink dispensed by the individual rollers by controlling the duty cycle, which is the ratio of the contact time to the control cycle τ / T1, where τ is the contact time between the individual rollers and the main roller, and T1 is the control period of the individual rollers. The target value of the ink supply amount at each individual roller (individual graph data) is defined as gr, the initial value of gr determined from the image to be printed is gri, the average value of the individual graph data gr across all duct rollers (average graph data) is g, and the initial value of g is gi. The control device is configured to control the duty cycle of each individual roller based on the individual graph data gr, and to change the individual graph data gr either by eliminating the error between the measured print density and the target print density, or by operator input. The duty cycle of the printing press is corrected by a correction device. In this invention, a correction device is used, Let ge be the stable value of the average graph data g, and gre be the stable value of individual graph data gr. Collect data including the initial values ​​gi and gri and the stable values ​​ge and gre of the graph data g and gr. Based on the difference between the distribution of the average graph data stability value ge and the distribution of the initial value gi in the collected data, the base parameter B is updated. When classifying the collected data by print speed range, the speed parameter V, which is a parameter for each print speed range, is updated based on the difference between the distribution of the average stable value ge of the graph data and the distribution of the initial value gi for each print speed range. When dividing the collected data into ranges of average graph data g, the area parameter F, which is a parameter for each range of average graph data g, is updated for each range of average graph data g based on the difference between the distribution of the stable value ge of the average graph data and the distribution of the initial value gi. The collected data is processed for each individual roller, and the roller parameter R, which is a parameter for each individual roller, is updated based on the difference between the distribution of the stable value gre and the distribution of the initial value gri in the individual graph data. Based on the updated base parameter B, the speed parameter V corresponding to the printing speed, and the area parameter F corresponding to the average graph data g, the duty cycle of all individual rollers is changed. Based on the updated roller parameter R, the duty cycle of the corresponding individual roller is changed. This invention provides the following effects and benefits. 1) Base parameter B corrects for overall errors such as ink effects and press conditions. This error is independent of errors that depend on printing speed, image area ratio, and individual rollers. 2) The speed parameter V is used to correct errors that depend on the printing speed. 3) The area parameter F is used to correct for errors that depend on the pattern area ratio. 4) The roller parameter R is used to correct for errors in each roller. 5) As a result of the above, fluctuations in the state of the printing press can be compensated for, and printing can start from a nearly optimal duty cycle. 6) Printing can be started at a nearly optimal duty cycle, resulting in less paper waste and enabling high-quality printing even for less experienced operators. 7) When printing on materials other than paper, such as cans or CD-ROMs, it is possible to reduce the loss of print quality until it stabilizes. The base parameter B applies to all individual rollers. The area parameter F uses the parameter of the category to which the average graph data g belongs. The speed parameter V uses the parameter of the category to which the printing speed belongs. The roller parameter R is a parameter specific to each roller. The stable value gre of the individual graph data gr is measured simultaneously with, for example, the stable value ge of the average graph data g. In this specification, the descriptions concerning the printing press also apply to the correction device and correction method. The difference between the distribution of stable values ​​and the distribution of initial values ​​is, for example, the difference between the mean value of the stable values ​​and the mean value of the initial values, or the difference between the median value of the stable values ​​and the median value of the initial values. Differences in distribution can be easily handled by the difference in means or the ratio of means, and the difference in means and the ratio of means represent essentially the same thing. The correction device is preferably, If the average value of the difference between the stable value and the initial value (ge-gi) in the collected data is positive, increase the base parameter B; if the average value of the difference between the stable value and the initial value (ge-gi) is negative, decrease the base parameter B. When classifying the collected data by print speed range, the speed parameter V for each print speed range is increased if the average value of the stable value and the initial difference ge-gi is positive, and decreased if the average value of the stable value and the initial difference ge-gi is negative. When dividing the collected data into ranges of average graph data g, the area parameter F, which is a parameter for each range of average graph data g, is increased if the average difference ge-gi between stability and initial value is positive, and decreased if the average difference ge-gi between stability and initial value is negative. The collected data is processed for each individual roller, and the roller parameter R, which is a parameter for each individual roller, is increased if the average of the difference between the stable value and the initial value of the individual graph data (gre-gri) is positive, and decreased if the average of the difference between the stable value and the initial value of the individual graph data (gre-gri) is negative. If the updated base parameter B, the speed parameter V corresponding to the printing speed, and the area parameter F corresponding to the average graph data g are greater than 1, the duty cycle of all individual rollers will be increased; if they are less than 1, the duty cycle of all individual rollers will be decreased. The system is configured to increase the duty cycle of the corresponding individual roller if the updated roller parameter R is greater than 1, and to decrease the duty cycle of the corresponding individual roller if it is less than 1. If two of the base parameter B, speed parameter V, and area parameter F are greater than 1 and the remaining parameter is less than 1, the duty cycle is controlled by a majority vote logic of B, V, and F. For example, the product of the three parameters (B, V, F) is compared to 1. If the product (B, V, F) is greater than 1, the duty cycle of all individual rollers is increased; if the product (B, V, F) is less than 1, the duty cycle of all individual rollers is decreased. The sign of ge-gi is the same as whether ge / gi is greater than 1 or whether gi / ge is less than 1. The average value is the average value of the collected data, and all data may be used, or unreliable data outside the center of the data may be excluded. In addition, when updating the speed parameter V, the collected data is divided based on the printing speed, and when updating the area parameter F, the collected data is divided based on the average graph data. Determining the initial value gri of individual graph data from the image to be printed means, for example, determining the initial value gri from the image area ratio for each individual duct roller. Preferably, the correction device processes the data as follows: For the three parameters, base parameter B, speed parameter V, and roller parameter F, only data where the average graph data g is greater than or equal to a predetermined value is evaluated from the collected data, and data below the predetermined value is not evaluated. In contrast, for the area parameter F, both data where the average graph data g is greater than or equal to a predetermined value and data below the predetermined value are evaluated. When the average graph data g is small, the print density becomes unstable, so by targeting only data where the average graph data g is greater than or equal to a predetermined value, the base parameter B, speed parameter V, and roller parameter F can be changed with high reliability. Furthermore, for the area parameter F, which needs to cover a wide density range, both data where the average graph data g is greater than or equal to a predetermined value and data below the predetermined value are evaluated. Preferably, the correction device is configured to change four parameters—base parameter B, velocity parameter V, area parameter F, and roller parameter R—to eliminate only a portion of the difference between the average graph data's stable value ge and its initial value gi, or the difference between the individual graph data's stable value gre and its initial value gri. In this way, the correction parameters asymptotically approach the optimal value through repeated updates, and the correction parameters do not oscillate. Preferably, the correction device is configured to correct the graph data ge, gi, gre, or gri when changing any of the base parameter B, speed parameter V, area parameter F, or roller parameter R, in order to compensate for the effect of the parameter change, and then change the other parameters based on the corrected graph data ge, gi, gre, or gri. In this way, errors that have already been processed by other parameters are not processed again with the new parameters, thus preventing overcorrection. Plan view showing the ink fountain, fountain roller, duct roller, and mixing roller. Waveform diagram showing the control waveform of the duct roller. Block diagram of the printing press in the embodiment. Diagram schematically showing the graph data file. Block diagram showing the correction device and surrounding printing units, etc., in the embodiment. Flowchart showing the base parameter update algorithm in the embodiment. Flowchart showing the speed parameter update algorithm in the embodiment. Flowchart showing the area parameter update algorithm in the embodiment. Flowchart showing the roller parameter update algorithm in the embodiment. The following are optimal embodiments for carrying out the present invention. These embodiments do not limit the scope of the present invention. The scope of the present invention is determined by taking into account the well-known art and understanding the claims in accordance with the understanding of those skilled in the art. Figures 1 to 9 show the correction device 20 and the correction method of the embodiment. As shown in Figure 1, ink is stored in the ink fountain 2, the fountain roller 4 contacts the ink fountain 2, and the ductor roller 6 draws ink from the fountain roller 4. The ductor roller 6 consists of a plurality of individual rollers 7, and the rollers 7 move back and forth in the direction of the arrows in Figure 1 between a position in contact with the fountain roller 4 and a position where they do not contact it, and are individually controlled. In this specification, when the term "ductor roller 6" is used, it refers to the entire plurality of rollers 7, and when simply referred to as "roller 7", it refers to an individual roller 7. 8 is a plurality of mixing rollers, of which only one is shown, and it mixes the ink, which is then supplied from the mixing roller 8 to the printing cylinder. Figure 2 shows the control waveform of roller 7. Roller 7 moves back and forth between a position where it contacts the inkwell roller 4 (on position) and a position where it does not contact it (off position) due to air pressure or the like. The control period of roller 7 is shown as T1, and the on time (contact time with the inkwell roller) is shown as τ. By controlling the on time τ, the amount of ink drawn out by each roller 7 is controlled. It is arbitrary whether to fix the period T1 and control the on time τ, fix τ and control T1, or control both τ and T1. Roller 7 draws ink from the inkwell roller 4, and the amount of ink drawn out can be controlled by controlling the duty cycle (ratio of contact time to control period τ / T1). Figure 3 shows a printing press 1, which is equipped with a printing unit 10 (hereinafter referred to as unit 10) for each ink such as CMYK, and also includes a paper feeder 11 and a paper discharger 12. For example, a densitometer 14 provided in the paper discharger 12 measures the print density of the printing paper. The print density is measured at each position corresponding to each roller 7 and input to a feedback device 15, which controls the amount of ink dispensed for each roller 7 by controlling the contact time τ. During this time, the graph data changes. The type of printing press 1 is arbitrary, and the unit 10 may be an inker of a can printing press, etc. It is also possible to have a printing press that does not have a densitometer and corrects individual graph data gr by visual inspection by the operator. In this case, instead of the feedback device 15, a duty cycle control device is provided, and the operator inputs individual graph data gr to the control device to eliminate the difference between the target print density and the actual print density. The correction device 20 outputs correction parameters to the feedback device 15. There are four types of correction parameters: base parameter B, which corrects for variations in print density (hereinafter referred to as "density") due to the type of ink and the state of unit 10; speed parameter V, which corrects for variations in density due to the printing speed; area parameter F, which corrects for variations in density due to the values ​​of the graph data; and roller parameter R, which corrects for variations in density due to the state of each roller 7. Parameters dealing with the properties of the printing paper may also be added. These parameters are meaningful for the combination of unit 10, printing paper, and ink. If printing has been done with this combination before, the initial values ​​of parameters B, V, F, and R are determined from the graph data at the time of printing. If printing has not been done before, the initial values ​​of parameters B, V, F, and R are set to 1, or the parameters B, V, F, and R from a similar combination are set as the initial values. Replacing the rollers in the printing unit 10, cleaning the water tank, etc., will significantly change the state of the printing unit 10. When the state of the printing unit 10 changes significantly, it is preferable to initialize the parameters B, V, F, and R. Figure 4 schematically shows a graph data file. The amount of ink that the duct roller should draw (target value of ink supply) is called graph data. When the plate is determined, that is, when the image to be printed is determined, the graph data gr for each roller 7 is determined from the image area ratio of the plate. Graph data gr is the target value of the ink draw amount and exists for each ink such as CMYK. The graph data file contains the average graph data g for the entire duct roller, the individual graph data gr for each roller, and the printing speed, etc. In addition, regarding the graph data g and gr, the recall values ​​(initial values) gi and gri of the graph data determined from the image area ratio, and the stable values ​​ge and gre of the graph data after feedback based on print density are also included. ge and gre are the values ​​when the print density is approaching the target value and stabilizing, for example, they are the graph data in the latter half of printing. ge and gre are sometimes called the graph data at the end of printing. The graph data file in Figure 4 is generated for each plate during the printing process. Figure 5 shows the correction device 20. The print density is measured by the densitometer 14 and compared with the target print density value (for example, the operator's input value) stored in the memory 16 by the comparator 17. The controller 18 controls the duct roller by changing the graph data g and gr to eliminate the density error. The memory 16, comparator 17, and controller 18 constitute the feedback device 15 shown in Figure 3. The correction device 20 is implemented by a suitable computer and is part of the printer 1. However, if multiple printers 1 are centrally controlled by a host computer via a LAN, the correction device 20 may be implemented by the host computer. The correction device 20 monitors changes in the graph data at the controller 18 and stores the graph data file shown in Figure 4 in the memory 21. The correction device 20 updates the correction parameters at the end of the workday and stores the changes in the values ​​of the correction parameters (for example, initial value and current value). The correction parameters are updated when the state of the printing unit 10 changes. The cumulative value of the changes in the correction parameters due to the updates represents the changes in the state of the printing unit 10. Therefore, displaying the cumulative value of the changes in the correction parameters due to updates on the display unit 32 can alert the operator to changes in the state of the printing unit 10. 22 is a means for updating the base parameter B, 24 is a means for updating the speed parameter V, 26 is a means for updating the area parameter F, and 28 is a means for updating the roller parameter R. The correction means 30 inputs the updated parameters to the controller 18, and the controller 18 corrects the ON duty cycle of the individual rollers 7 by the product kr of these parameters. Figures 6 to 9 show the parameter update algorithm. In step 1 of Figure 6, graph data files are collected, which means storing the files in memory 21. First, the base parameters that reflect the ink type and the state of unit 10 are updated. If the graph data is below a predetermined value, the print density tends to become unstable, so graph data files that are above the predetermined value are extracted (step 2). Here, the graph data can be either the initial value gi (value at the time of recall) or the stable value ge (value at the time of stable printing before the end of printing). Verify that there are more than a specified number of sorted files (for example, two or more). Furthermore, Assuming d = ge / gi, confirm that d is not distributed symmetrically around 1, but rather is biased towards either the greater or lesser side of 1 (steps 3 and 4). d is the percentage correction applied by the feedback device 15 to the graph data at the time of recall (graph data at the start of printing). d > 1 indicates an increase in graph data, and d < 1 indicates an increase in graph data. Also, d is a value per file. If the number of files is small, the data reliability is low, and if d is distributed symmetrically around 1, there is no need to update the base parameter. Confirmation that the distribution of d is not symmetrical around 1 can be omitted. If there are multiple files with graph data exceeding a predetermined value, and the distribution of d is biased towards either being greater than or less than 1, the base parameter B is updated. Let A(d) be the mean of d. Let Bnew = Bold × (1 + (A(d) - 1)a) (Step 5). Here, a is the correction rate, where 0 < a < 1, Bold is the base parameter before updating, and Bnew is the base parameter after updating. By correcting only a part of the error in the base parameter B without completely eliminating it, the base parameter B is asymptotically approached to an appropriate value through multiple updates. The change due to the update is determined by (A(d) - 1)a, and an upper limit may be set on the absolute value of (A(d) - 1)a. After updating B, the value of ge is updated to update other parameters such as the velocity parameter on the graph data in which B has already been updated. Substitute ge2 = ge / (1 + (A(d) - 1)a) (Step 6). Figure 7 shows the update of the speed parameter V, and files with graph data above a predetermined value are targeted, in other words, files with graph data that are too low and therefore prone to unstable print density are excluded, and the files are sorted by print speed range (step 11). For speed ranges where there are multiple valid files (number of files with graph data above a predetermined value), for each file Calculate d2 = ge2 / gi (Step 12), and calculate the average of d2 for each speed range. The value of the average of d2 for each speed range that is closest to 1 is set as D. The speed range to which D belongs is set as the reference speed range, and the speed parameters are not updated in the reference speed range, while in the other speed ranges the average is divided by D to obtain D2 (Step 14). In the speed range-based update, it is assumed here that the change due to the update should be 0 in one of the speed ranges. This assumption may be omitted. In addition, a constraint is added that D2 changes smoothly, and starting from the reference speed range, the value of D2 is restricted to being midway between the values ​​in the left and right speed ranges. If this constraint is violated, in the corresponding speed range... Let D2 = 1. The velocity parameter is updated in the same way as for base parameter B (step 15), and an upper limit may be set on the change due to the update. Also, the value of ge2 is set for the next area parameter update. Substitute ge3 = ge2 / (1 + (D2 - 1)b) (Step 16). Figure 8 shows the updating of the area parameter F. In updating the area parameter F, the parameter F is updated for a wide range of average graph data g, so files with graph data below a predetermined value are also included. That is, the area parameter is divided into ranges of average graph data g over a wide range, and the files (valid files) within each range (range of graph data g) are sorted (step 21). For area ranges (range of graph data g) where multiple valid files exist, the ratio d3 = ge3 / gi is calculated (step 22). The average of d3 in the area range where the average value of d3 is closest to 1 is taken as E, and since E is used as the reference, the average in the other area ranges is divided by E to obtain E2 (step 24). Also, the parameter F is not updated in the area range corresponding to E. This is based on the assumption that since the update is for each area range (range of graph data), the change due to the update should be 0 in at least one of the area ranges. It is also assumed that E2 changes smoothly from 1 in the reference area range, and that the value of E2 should be midway between the values ​​in the area ranges on both sides. If this assumption is violated, for example, replace E2 with 1. Then, similar to the case of the base parameter B etc., the area parameter F is updated in step 25. In step 26, as preparation for updating the roller parameter, ge3 is replaced with ge4 by ge4 = ge3 / (1 + (E2 - 1)f). Regarding the update of the area parameter, if the update is delayed due to insufficient number of files, it may be processed assuming that there is a file of graph data for each roller 7. The file of graph data describes the call value gri of the graph data for each roller and the value gre at the end. When updating the parameters B, V, F, similar to the graph data ge, the graph data gre is also replaced in steps 6, 16, 26 of FIGS. 6 to 8. Thereby, the influence of updating the base parameter, speed parameter, and area parameter is corrected. In FIG. 9, the parameter R for each roller is updated. In step 31, the files in which the graph data gr for each individual roller is greater than or equal to a predetermined value are sorted. In step 32, it is checked whether d4 = gre / gri (gre has been replaced with a new value in steps 6, 16, 26) is biased from 1 in distribution. When the distribution is biased from 1 (step 33), the parameter for each roller is updated in the same manner as the update of the base parameter B (step 34). Then, the new parameter is output to the correction means 30 and stored by the correction means 30 (step 35). In the above processing, the update of the base parameter B is performed first, the update of the roller parameter R is performed last, and the order of updating the area parameter F and the speed parameter V is arbitrary. In the embodiment, the correction parameter is optimized by multiple updates. In other words, conditions are imposed on the update so that the correction parameter does not oscillate because the correction parameter is updated largely or is updated based on uncertain data. For example ・ The existence of a plurality of valid files, ・ The graph data is not less than a predetermined value (parameters B, V, R), ・ The correction rate is a number between 0 and 1, ・ An upper limit is set for the absolute value of the change due to the update, The parameters (parameters V, F) change smoothly according to the velocity range and area range. If parameter oscillation is not a concern, these conditions may be omitted. Important conditions for updates include not using files with graph data below a predetermined value for the base parameter B, speed parameter V, and roller parameter R, and not updating if there are no multiple valid files for each parameter B, V, F, and R. It is also important to asymptotically approach the optimal value through multiple updates by setting the correction rate to be greater than 0 and less than 1, or by imposing an upper limit on the change in parameters. Returning to Figure 5, the correction of the duty cycle for each individual roller will be explained. The controller 18 stores the printing speed, area ratio, or graph data g. The correction means 30 selects a speed category for the speed parameter V according to the printing speed, and an area category (area ratio category or graph data category) for the area parameter F according to the area ratio or graph data g, and reads out the corresponding speed parameter V and area parameter F. It also multiplies the parameters B, F, V, and R to obtain the product kr = B・V・F・R, and outputs kr to the controller 18. The controller 18 corrects the duty cycle for each individual roller 7, which is determined from the graph data gr, by multiplying it by kr, and controls the individual roller 7. Alternatively, instead of multiplying the duty cycle by kr, kr may be multiplied by the initial value gri of gr. In the embodiment, four parameters were multiplied, but the duty cycle correction coefficient is not limited to multiplication; it is sufficient for the duty cycle correction coefficient to be determined as a function of the four parameters. The four parameters can be updated independently; for example, if the area parameter F is not updated due to insufficient data, the other three parameters can be updated. In the embodiment, when the printing paper or ink is changed, the correction parameters from before the change are not used, but the parameters from before the change may be used. For example, the speed parameter V, which corrects the dependence on the printing press speed, and the roller parameter R, which corrects the dependence on individual duct rollers, may remain the same correction parameters even if the printing paper or ink is changed. In the embodiment, the feedback device 15 learns how to change the graph data and determines correction parameters. In the embodiment, the following effects can be obtained. 1) The base parameter B corrects the overall error that is independent of each individual roller, such as the influence of ink and the state of the printing machine, including printing speed, pattern area ratio. 2) The speed parameter V corrects the error that depends on the printing speed. 3) The area parameter F corrects the error that depends on the pattern area ratio. 4) The roller parameter R corrects the error for each roller. 5) By these, the state variation of the printing machine is corrected, and printing can start from an almost appropriate duty ratio. 6) Since printing can start from an almost appropriate duty ratio, waste paper is less, and high-quality printing can be achieved even without a skilled operator. 7) When printing on cans, CD-ROMs, etc. instead of paper, the loss until the printing density stabilizes can be reduced. 8) The correction parameters are updated based on reliable data so as to asymptotically approach the optimal values. Therefore, the correction parameters do not oscillate. The parameters B, V, F, and R are determined for each combination of the printing unit, paper, and type of ink. When changing the printing unit, changing the paper, or changing the ink, the file of the graph data in FIG. 4 may not be stored. The method for determining the initial values of the parameters B, V, F, and R in this case will be described. When the file of the graph data is not stored, ・ The parameters B, V, F, and R when using the same ink and paper with a similar printing unit, ・ The parameters B, V, F, and R when using paper with a similar paper quality, the same ink, and the same printing unit, ・ The parameters B, V, F, and R when using ink with a similar ink performance (a value empirically indicating the level of printing density with the same ink supply amount), the same printing unit, and the same paper, These can also be used as initial values ​​for parameters B, V, F, and R. That is, if one of the three elements that affect parameters B, V, F, and R is changed, the parameter values ​​when the other two elements remain the same can also be used as the initial values ​​for the parameters. In practical terms, spot inks other than CMYK pose a particular problem. Due to their diverse types, it is difficult to determine appropriate initial values ​​for parameters B, V, F, and R, and because they are used infrequently, parameter updates cannot be expected. Therefore, parameters V, F, and R are set to the values ​​from the previous application for a different ink, for example. In many cases, the ink adhesion for spot inks can be empirically evaluated. The spot parameter s is set as an empirical value indicating the percentage by which the ink supply is increased depending on the type of ink, with s=1 being the standard value, and a larger s indicating a greater increase in ink supply. The spot parameters for the newly used spot ink and the other ink used immediately before are used. For example, let B be the base parameter from the previous application, s' be the spot parameter of the previously used ink, and s be the spot parameter of the newly used spot ink, and s / s' × B be used as the initial value of the base parameter B for the newly used spot ink. 1. Printing press 2. Ink fountain 4. Fountain roller 6. Duct roller 7 Roller 8 Mixing roller 10 Unit 11 Paper feed device 12 Paper output device 14 Densitometer 15 Feedback device 16 Memory 17 Comparator 18 Controller 20 Correction device 21 Memory 22 Base parameter update means 24 Speed ​​parameter update means 26 Area parameter update means 28 Roller parameter update means 30 Correction means 32 Display unit T1: Period τ: On-time g: Graph data d, d2, d3, d4: Ratio of stable values ​​in the graph data to values ​​at the time of recall. B: Base parameter V: Speed ​​parameter F: Area parameter R: Roller parameter A(d): Average of d A(d4): Average of d4​ D2: The ratio of the average d2 for each speed range to the average at the base speed. E2: Ratio of the average d3 for each range of graph data to the average for the reference graph data. a,b,f,r: Correction rate