A method for adaptive tension adjustment and correction compensation in carton printing line
By using a paperboard parameter library and online identification technology, the tension adaptive adjustment and correction compensation of the carton printing line are realized, which solves the coupling disturbance problem between tension and correction control and improves the registration accuracy and yield.
Patent Information
- Application Number
- CN202611550090.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-10-08
- Publication Date
- 2026-11-17
AI Technical Summary
In existing carton printing lines, tension control and web correction control are designed independently, resulting in severe coupling disturbances. This makes it difficult to adapt to the time-varying equivalent tensile stiffness of the cardboard, affecting registration accuracy and yield.
By establishing a paperboard parameter library and using online identification technology to obtain paperboard characteristic information, an adaptive adjustment and correction compensation method is constructed, including parameter initialization, online identification, controller self-tuning, operation data acquisition, adaptive tension adjustment, correction compensation, and decoupling coordination, thereby achieving decoupled control of tension and correction loop.
It significantly shortens setup time, reduces scrap rate, improves registration accuracy and yield, avoids tension fluctuations and correction impacts, and achieves stable control for different types of paperboard.
Smart Images

Figure CN123105364A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of automatic control technology for printing and packaging equipment, specifically relating to a method for adaptive tension adjustment and correction compensation of a carton printing line. Background Technology
[0002] A carton printing line (printing-slotting-die-cutting-gluing line) is typically set up with a paper feeding unit, a multi-color printing unit, a slotting and creasing unit, a die-cutting unit, and a gluing unit in sequence. The paperboard is conveyed between each unit by cascaded actuators such as paper feeding rollers and traction rollers. It runs at high speed in a continuous or single-sheet manner, with the operating speed generally ranging from tens of meters to three hundred meters per minute. Due to frequent start-stop and speed changes for order changes and troubleshooting, the line is frequently started, stopped, and changed.
[0003] Maintaining appropriate and stable tension on the cardboard between units is a prerequisite for ensuring registration accuracy, slotting and die-cutting alignment accuracy, and preventing cardboard stretching, deformation, wrinkles, and even paper breaks. Simultaneously, the degree of cardboard alignment in the machine's width direction (lateral) directly affects quality indicators such as printing gaps, image misalignment, and slotting misalignment. Therefore, tension adjustment and lateral correction are two core components of the line control system.
[0004] In existing production line control systems, the tension loop and the correction loop are generally designed, tuned, and operated independently. However, the controlled objects of these two loops are physically coupled in two directions: on the one hand, the oscillation of the correction roller changes the wrapping length of the cardboard, directly disturbing the tension in that section; on the other hand, tension changes cause uneven lateral expansion and contraction of the cardboard, causing the edge position of the cardboard to shift. When controlled independently, the two loops mutually excite each other—adjusting the tension exacerbates the deviation, and the correction action, in turn, disturbs the tension, leading to continuous oscillations in severe cases. This necessitates either reduced speed or manual intervention, limiting the improvement of production line speed and yield.
[0005] In summary, the existing technology has at least the following technical problems that urgently need to be solved: First, the tension control parameters cannot adapt to the time-varying equivalent tensile stiffness of the cardboard, resulting in long machine changeover times and high scrap rates; Second, the web correction control cannot adapt to the time-varying detection delay with the machine speed, making it difficult to balance stability and web correction accuracy across the entire speed range; Third, the coupling disturbance between the tension loop and the web correction loop lacks decoupling methods, causing mutual interference between the two loops; Fourth, the lack of linkage between registration compensation and tension adjustment results in poor registration accuracy under dynamic operating conditions. Summary of the Invention
[0006] A method for adaptive tension adjustment and correction compensation in a carton printing line includes the following steps: S1. Parameter Initialization: Establish a paperboard parameter library. The paperboard parameter library uses paperboard feature information as an index to record paperboard feature information and corresponding historical identification parameters and historical controller parameters. The historical identification parameters include the equivalent tensile stiffness, static gain, and converted inertia obtained from previous online identifications. The historical controller parameters include the tension controller parameters obtained from previous self-tuning. After obtaining the paperboard feature information of the current order, search for matching records in the paperboard parameter library. If a match is found, use the historical identification parameters in the matching record as the initial value for this online identification, so that the identification starts from a point close to the true value, significantly shortening the excitation and convergence time and reducing paperboard waste in the paper feeding stage. If a match is not found, estimate the initial value of the equivalent tensile stiffness of the current paperboard according to the empirical model of paperboard structural mechanics, as the starting point for identification iteration, to ensure that the identification can also converge quickly when the new variety is first put on the machine. S2. Online Identification: During the paper feeding stage (when the machine speed is low and printing has not yet begun, a small excitation is allowed without affecting product quality), an excitation signal is superimposed on the speed difference between adjacent traction rollers to fully excite the tension channel, collect the tension response, and use the recursive least squares method with a forgetting factor to identify the equivalent tensile stiffness and static gain of the current paperboard online. Based on the tension response under the known acceleration command, the inertia is identified, that is, the inertial component in phase with the acceleration is separated from the tension response, and the inertial tension disturbance during the acceleration process is transformed into a deterministic component that can be accurately fed forward and compensated. S3. Controller self-tuning: Based on the equivalent tensile stiffness, the static gain of the object, and the converted inertia, the parameters of the tension controller are self-tuned according to the desired closed-loop characteristics, so that the dynamic quality (bandwidth, damping, overshoot) of the tension closed loop remains consistent under different paperboard types and different working conditions, eliminating the need for manual trial and error tuning, and realizing "identification and tuning, type change and adaptation". S4. Operational data acquisition: Real-time acquisition of tension in each section, linear speed of each traction roller, edge position of the cardboard and registration mark, and calculation of the time-varying edge detection delay based on the real-time linear speed. Since there is an installation distance between the edge sensor and the correction execution point, the edge position signal needs to be delayed before it acts on the correction point. This delay changes continuously with the vehicle speed and is a key time-varying parameter for correction loop tuning and compensation. S5. Tension Adaptive Adjustment: The acceleration inertia feedforward, the cascade speed difference feedforward, and the feedback control quantity generated by the tension controller are combined into a tension control quantity, and the speed difference command of each traction roller is output, forming a composite control structure of "feedforward bearing dynamic load and feedback eliminating residual error"; wherein, the acceleration inertia feedforward is determined by the converted inertia and real-time acceleration, and the coefficient of the cascade speed difference feedforward is refreshed according to the online identification result of the equivalent tensile stiffness; S6. Correction compensation: For the pure delay in the correction loop that varies with vehicle speed, a Smith predictor is constructed. The internal model delay parameter of the Smith predictor is refreshed online with the vehicle speed according to the edge detection delay. The correction deviation after Smith prediction compensation (i.e., the equivalent instantaneous deviation after deducting the delay) and the correction feedback control together generate a correction command, which is output to the correction execution mechanism. S7. Decoupling and Coordination: Estimate the tension disturbance caused by the correction action based on the equivalent tensile stiffness, and feed the tension disturbance forward to the tension adjustment channel to cancel it out; estimate the lateral offset caused by the tension change based on the pre-calibrated tension-offset coupling coefficient, and feed the lateral offset forward to the correction channel to cancel it out, thereby realizing the approximate dynamic decoupling of the tension loop and the correction loop, so that the two loops can be independently tuned as a single variable system and do not interfere with each other.
[0007] Furthermore, a method for adaptive tension adjustment and correction compensation in a carton printing line is provided. In step S1, the cardboard feature information of the current order is obtained, and a matching record is retrieved from the cardboard parameter library, as follows: The cardboard characteristics include flute type, basis weight, width, and moisture content. Flute type and basis weight determine the mechanical properties of the cardboard cross-section, width determines the load-bearing cross-sectional area, and moisture content has a significant softening effect on stiffness. The retrieval is a fuzzy retrieval, with matching conditions of identical flute types, basis weight difference less than 10%, and moisture content difference less than 2%. Relaxing the matching precision improves the parameter database hit rate and reuse value. When a matching record is found, because the initial value is close to the true value, the duration of the excitation signal in step S2 is halved, shortening the time. Shorten the paper feeding time and reduce tension fluctuations caused by excitation; if a miss occurs, estimate the initial value of the equivalent tensile stiffness according to the flute type, face paper weight and liner paper weight, width and moisture content correction coefficient, that is, look up the cross-sectional equivalent coefficient according to the flute type, calculate the reference stiffness according to the face paper and liner paper weight and width, and then multiply it by the moisture content correction coefficient; wherein, for every 1% increase in moisture content, the initial value of the equivalent tensile stiffness is reduced by 3% to 5% to reflect the weakening of fiber bonding force by moisture and ensure that the initial value falls within the convergence region of online identification.
[0008] Furthermore, a method for adaptive tension adjustment and correction compensation in a carton printing line is provided. In step S2, during the paper feeding stage, an excitation signal is superimposed on the speed difference between adjacent traction rollers, and the tension response is collected. The equivalent tensile stiffness and static gain of the current paperboard are identified online using a recursive least squares method with a forgetting factor, as detailed below: The excitation signal is a pseudo-random binary sequence and a bidirectional step signal superimposed on the speed difference between adjacent traction rollers, with an amplitude of 1% to 2% of the current linear velocity; The discrete model used for identification is T(k) = a·T(k-1) + b·δv(k-1-d) + c, where T(k) is the tension of the k-th beat, δv(·) is the speed difference, d is the delay beat number, and c is the steady-state bias. The forgetting factor of the recursive least squares method with forgetting factor is taken as 0.95~0.99, taking into account both convergence speed and parameter stability (if the factor is too small, noise will be amplified and parameters will jump; if it is too large, tracking will be slow). When the relative change of parameter increment is less than 1% for no less than 100 beats, the identification is considered to have converged, the parameters are locked, and the excitation signal is canceled. After the identification has converged, EA = b·L / Ts and K_T = b / (1-a) are used. The equivalent tensile stiffness and the static gain of the object are converted, where L is the length of the section between adjacent traction rollers and Ts is the control cycle; the identification method of the converted inertia is as follows: execute an acceleration command with a known slope, extract the component in phase with the acceleration in the tension response, and use the ratio of this component to the acceleration command as the equivalent inertia converted to the roller surface. The separation and identification of inertia and stiffness are achieved by using the proportional relationship between acceleration and inertial tension, thus avoiding parameter coupling.
[0009] Furthermore, a method for adaptive tension adjustment and correction compensation in a carton printing line is provided. In step S3, the desired closed-loop characteristic is a second-order characteristic with a damping ratio of 0.7 to 0.9. The closed-loop bandwidth is determined according to the constraint of the edge detection delay and is graded according to the vehicle speed. The tension controller parameters are obtained by pole configuration, including Kp=ωc / K_T and Ki=ωc² / (2·K_T), where ωc is the closed-loop bandwidth and K_T is the static gain of the object. Since K_T comes from online identification, the controller parameters are automatically refreshed after the cardboard is changed or the parameters drift, and the closed-loop characteristics remain consistent across different varieties.
[0010] Furthermore, a method for adaptive tension adjustment and correction compensation in a carton printing line is provided. Step S4 also includes tension measurement redundancy cross-calibration: the tension estimate obtained by the servo motor torque soft measurement is cross-calibrated with the measurement value of the tension sensor. When the deviation between the two exceeds 10% and continues for more than a preset time, an alarm is triggered and control is switched to the tension estimate. The edge detection delay is calculated in real time according to τ=Ls / v and discretized according to the control cycle, where Ls is the distance from the correction point to the edge sensor and v is the real-time linear velocity. Since v changes continuously during acceleration and deceleration, the delay parameter is refreshed frame by frame and rounded to the nearest control cycle number to ensure that the internal model delay of the Smith predictor is always synchronized with the actual delay.
[0011] Furthermore, a method for adaptive tension adjustment and correction compensation in a carton printing line is provided. In step S5, the feedback control quantity is generated by an incremental PID algorithm. During operation, the PID parameters are finely adjusted online according to the tension error and its rate of change based on fuzzy rules. Amplitude limiting and output rate of change limit are applied to the tension control quantity, and anti-saturation processing is implemented on the integral term. Step S5 further includes disturbance observation compensation: the tension measurement value is processed through the inverse model of the controlled object in the tension section to obtain the inverse model output, the difference between the inverse model output and the tension control quantity is processed through low-pass filtering to obtain the disturbance estimate (the low-pass filtering is used to suppress the measurement noise amplified by the inverse model differentiation operation), and the disturbance estimate is used as a compensation quantity in the synthesis to offset the load disturbance and friction disturbance; Step S5 also includes eccentricity compensation: the roller rotation frequency is calculated according to the real-time rotation speed of each traction roller, and the center frequency of the digital notch filter follows the rotation frequency in real time to suppress the periodic tension fluctuations of the rotation frequency and its harmonics caused by roller eccentricity. Since the rotation frequency changes continuously with the vehicle speed, the fixed frequency notch filter cannot cover it. The real-time tracking center frequency makes the notch filter always aligned with the eccentricity disturbance frequency during the speed change process, and a notch filter can be connected in parallel at the harmonic to suppress the secondary component.
[0012] Furthermore, a method for adaptive tension adjustment and correction compensation in a carton printing line is provided. In step S6, the correction deviation, after being estimated and compensated by Smith, and the correction feedback control together generate a correction command, which is output to the correction actuator. The correction command is divided into the following cases: When the correction actuator is a servo correction roller set in the continuous conveying section, the correction command is a swing angle command, the swing angle is limited to ±2° and the rate of change is limited to no more than 1° / s, to prevent excessive wrap angle changes from causing sudden tension changes, paperboard slippage or even paper breakage. When the correction actuator is the left and right paper feed rollers of a single sheet conveying section, the correction command is the differential speed command or differential pressure command of the left and right paper feed rollers; when the correction actuator is the left and right paper feed rollers, it also includes a skew correction step: the edge position is detected at two points before and after the paperboard conveying direction, the skew angle is obtained by the ratio of the difference between the two edge positions to the distance between the two detection points, the skew angle is converted into the additional differential speed of the left and right paper feed rollers, and then output after being superimposed with the deviation correction amount, so as to correct the deviation and the skew of the paperboard at the same time, and ensure the accuracy of subsequent printing registration and die-cutting.
[0013] Furthermore, a method for adaptive tension adjustment and correction compensation in a carton printing line is provided. In step S7, the tension disturbance caused by the correction action is estimated based on the equivalent tensile stiffness. The tension disturbance caused by the correction action is calculated as ΔTθ=EA·ΔL(θ) / L, where ΔL(θ)=2·Lt·sin²(θ / 2), where θ is the swing angle of the correction roller, Lt is the front and rear wrapping length of the correction roller, L is the segment length, and EA is the equivalent tensile stiffness. This tension disturbance is then superimposed on the tension setpoint. When the correction is expected to increase (decrease) the tension, the tension setpoint is pre-decreased (increased) accordingly to protect the tension ring from the correction impact. The lateral offset caused by the tension change is calculated as Δy=β·ΔT The calculation is performed, where β is the pre-calibrated tension-offset coupling coefficient, ΔT is the tension change, and the lateral offset is superimposed on the input of the correction loop; it also includes priority arbitration: when the correction action amplitude exceeds the first preset threshold, the integral action of the tension controller is paused for a preset duration; when the linear acceleration exceeds the second preset threshold, the rate of change limit of the correction command is tightened; and the closed-loop bandwidths of the tension loop and the correction loop are staggered by no less than 3 times to avoid mutual excitation and resonant coupling caused by the overlap of the two loop frequency bands.
[0014] The beneficial effects of this invention are as follows: Initial values are provided for online identification through a paperboard parameter library retrieval. During the paper feeding stage, the equivalent tensile stiffness, static gain, and reduced inertia of the identified paperboard are superimposed. When switching to paperboards with different flute types, basis weights, and moisture contents, the object parameters are automatically acquired without manual readjustment, resulting in shorter setup time and less waste during trial runs. The tension controller parameters are self-tuned according to the identification results and the desired closed-loop characteristics. The bandwidth and damping of the tension closed loop remain consistent across different paperboard types, avoiding mismatch oscillations or slow response after type switching with fixed parameter control. The acceleration-inertia feedforward is determined by the reduced inertia and real-time acceleration. The cascaded speed difference feedforward coefficient is updated online with the equivalent tensile stiffness. Tension disturbances during acceleration / deceleration and between sections are canceled before feedback adjustment, significantly reducing the amplitude of tension fluctuations. Disturbance observation compensation cancels load and friction disturbances. The center frequency of the digital notch filter follows the roller rotation frequency in real time, suppressing periodic tension fluctuations caused by eccentricity, and preventing mismatch during variable speed operation. The Smith predictor's internal model delay is updated online based on edge detection delay and vehicle speed. The correction loop balances speed and stability at both low and high speeds, eliminating the need for separate tuning for each speed. A two-loop bidirectional decoupling is achieved between the correction-tension disturbance feedforward based on equivalent tensile stiffness and the offset feedforward based on the tension-offset coupling coefficient. This ensures that tension is unaffected during large-scale corrections and that registration drift is not caused during tension adjustments. Both loops can each utilize higher bandwidths. Redundant cross-calibration between servo torque soft measurement and the tension sensor provides an alarm and seamless degradation to soft measurement control in case of measurement anomalies, preventing paper breaks and unplanned downtime. Attached Figure Description
[0015] Figure 1 A flowchart of a method for adaptive tension adjustment and correction compensation in a carton printing line; Detailed Implementation A method for adaptive tension adjustment and correction compensation in a carton printing line, the specific process of which is as follows: Figure 1 As shown, it includes the following steps: S1. Parameter Initialization: Establish a paperboard parameter library. The paperboard parameter library uses paperboard feature information as an index to record paperboard feature information and corresponding historical identification parameters and historical controller parameters. The historical identification parameters include the equivalent tensile stiffness, static gain, and converted inertia obtained from previous online identifications. The historical controller parameters include the tension controller parameters obtained from previous self-tuning. After obtaining the paperboard feature information of the current order, search for matching records in the paperboard parameter library. If a match is found, use the historical identification parameters in the matching record as the initial value for this online identification, so that the identification starts from a point close to the true value, significantly shortening the excitation and convergence time and reducing paperboard waste in the paper feeding stage. If a match is not found, estimate the initial value of the equivalent tensile stiffness of the current paperboard according to the empirical model of paperboard structural mechanics, as the starting point for identification iteration, to ensure that the identification can also converge quickly when the new variety is first put on the machine. S2. Online Identification: During the paper feeding stage (when the machine speed is low and printing has not yet begun, a small excitation is allowed without affecting product quality), an excitation signal is superimposed on the speed difference between adjacent traction rollers to fully excite the tension channel, collect the tension response, and use the recursive least squares method with a forgetting factor to identify the equivalent tensile stiffness and static gain of the current paperboard online. Based on the tension response under the known acceleration command, the inertia is identified, that is, the inertial component in phase with the acceleration is separated from the tension response, and the inertial tension disturbance during the acceleration process is transformed into a deterministic component that can be accurately fed forward and compensated. S3. Controller self-tuning: Based on the equivalent tensile stiffness, the static gain of the object, and the converted inertia, the parameters of the tension controller are self-tuned according to the desired closed-loop characteristics, so that the dynamic quality (bandwidth, damping, overshoot) of the tension closed loop remains consistent under different paperboard types and different working conditions, eliminating the need for manual trial and error tuning, and realizing "identification and tuning, type change and adaptation". S4. Operational Data Acquisition: Real-time acquisition of tension in each section, linear speed of each traction roller, edge position of the cardboard, and registration marks, providing a multi-source measurement basis for tension adjustment, correction compensation, and registration monitoring. The edge detection delay, which varies with vehicle speed, is calculated based on the real-time linear speed. Since there is an installation distance between the edge sensor and the correction execution point, the edge position signal needs to have a pure time delay before it acts on the correction point. This delay changes continuously with vehicle speed and is a key time-varying parameter for correction loop tuning and compensation. S5. Tension Adaptive Adjustment: The acceleration inertia feedforward, the cascaded speed difference feedforward, and the feedback control quantity generated by the tension controller are combined into a tension control quantity, and the speed difference command of each traction roller is output, forming a composite control structure of "feedforward bearing dynamic load and feedback eliminating residual error"; wherein, the acceleration inertia feedforward is determined by the converted inertia and real-time acceleration, and actively compensates for tension drop or surge caused by inertia during acceleration, deceleration, and jogging; the cascaded speed difference feedforward is used to compensate for the disturbance of tension in this section caused by the speed adjustment of the traction rollers in the adjacent section through the paperboard path, and its feedforward coefficient is refreshed with the online identification result of the equivalent tensile stiffness, and the compensation strength automatically follows the change of object characteristics after the paperboard type is changed or the parameters drift; S6. Correction Compensation: To address the pure delay in the correction loop that varies with vehicle speed, a Smith predictor is constructed. The pure lag element is approximately eliminated from the closed-loop characteristic equation using the internal model delay, overcoming the correction lag, overshoot, and oscillation caused by the delay. The internal model delay parameter of the Smith predictor is refreshed online with the vehicle speed according to the edge detection delay. The correction deviation after Smith prediction compensation (i.e., the equivalent instantaneous deviation after deducting the delay) and the correction feedback control together generate a correction command, which is output to the correction execution mechanism. S7. Decoupling and Coordination: Estimate the tension disturbance caused by the correction action based on the equivalent tensile stiffness, and feed the tension disturbance forward to the tension adjustment channel to cancel it out; estimate the lateral offset caused by the tension change based on the pre-calibrated tension-offset coupling coefficient, and feed the lateral offset forward to the correction channel to cancel it out, thereby realizing the approximate dynamic decoupling of the tension loop and the correction loop, so that the two loops can be independently tuned as a single variable system and do not interfere with each other.
[0016] Furthermore, a method for adaptive tension adjustment and correction compensation in a carton printing line is provided. In step S1, the cardboard feature information of the current order is obtained, and a matching record is retrieved from the cardboard parameter library, as follows: The cardboard characteristics include flute type, basis weight, width, and moisture content. Flute type and basis weight determine the mechanical properties of the cardboard cross-section, width determines the load-bearing cross-sectional area, and moisture content has a significant softening effect on stiffness. The retrieval is a fuzzy retrieval, with matching conditions of identical flute types, basis weight difference less than 10%, and moisture content difference less than 2%. Relaxing the matching precision improves the parameter database hit rate and reuse value. When a matching record is found, because the initial value is close to the true value, the duration of the excitation signal in step S2 is halved, shortening the time. Shorten the paper feeding time and reduce tension fluctuations caused by excitation; if a miss occurs, estimate the initial value of the equivalent tensile stiffness according to the flute type, face paper weight and liner paper weight, width and moisture content correction coefficient, that is, look up the cross-sectional equivalent coefficient according to the flute type, calculate the reference stiffness according to the face paper and liner paper weight and width, and then multiply it by the moisture content correction coefficient; wherein, for every 1% increase in moisture content, the initial value of the equivalent tensile stiffness is reduced by 3% to 5% to reflect the weakening of fiber bonding force by moisture and ensure that the initial value falls within the convergence region of online identification.
[0017] Furthermore, a method for adaptive tension adjustment and correction compensation in a carton printing line is provided. In step S2, during the paper feeding stage, an excitation signal is superimposed on the speed difference between adjacent traction rollers, and the tension response is collected. The equivalent tensile stiffness and static gain of the current paperboard are identified online using a recursive least squares method with a forgetting factor, as detailed below: The excitation signal is a pseudo-random binary sequence and a bidirectional step signal superimposed on the speed difference between adjacent traction rollers. The amplitude is 1% to 2% of the current linear speed. If the amplitude is too small, the signal-to-noise ratio will be insufficient and the identification accuracy will be poor. If the amplitude is too large, the tension fluctuation will exceed the limit and may damage the cardboard or break the paper. The 1% to 2% amplitude balances the excitation sufficiency and the operational safety. The pseudo-random binary sequence has a wide spectrum, which can excite the object in various frequency bands in a short time. The bidirectional step can overcome the asymmetry caused by nonlinearity such as static friction and gap. The discrete model used for identification is T(k) = a·T(k-1) + b·δv(k-1-d) + c, where T(k) is the tension of the k-th beat, δv(·) is the speed difference, d is the delay beat number, and c is the steady-state bias. The forgetting factor of the recursive least squares method with forgetting factor is taken as 0.95~0.99, taking into account both convergence speed and parameter stability (if the factor is too small, noise will be amplified and parameters will jump; if it is too large, tracking will be slow). When the relative change of parameter increment is less than 1% for no less than 100 beats, the identification is considered to have converged, the parameters are locked, and the excitation signal is canceled. After the identification has converged, EA = b·L / Ts and K_T = b / (1-a) are used. The equivalent tensile stiffness and the static gain of the object are converted, where L is the length of the section between adjacent traction rollers and Ts is the control cycle; the identification method of the converted inertia is as follows: execute an acceleration command with a known slope, extract the component in phase with the acceleration in the tension response, and use the ratio of this component to the acceleration command as the equivalent inertia converted to the roller surface. The separation and identification of inertia and stiffness are achieved by using the proportional relationship between acceleration and inertial tension, thus avoiding parameter coupling.
[0018] Furthermore, a method for adaptive tension adjustment and correction compensation in a carton printing line is provided. In step S3, the desired closed-loop characteristic is a second-order characteristic with a damping ratio of 0.7 to 0.9. The closed-loop bandwidth is determined according to the constraint of the edge detection delay and is graded according to the vehicle speed. The tension controller parameters are obtained by pole configuration, including Kp=ωc / K_T and Ki=ωc² / (2·K_T), where ωc is the closed-loop bandwidth and K_T is the static gain of the object. Since K_T comes from online identification, the controller parameters are automatically refreshed after the cardboard is changed or the parameters drift, and the closed-loop characteristics remain consistent across different varieties.
[0019] Furthermore, a method for adaptive tension adjustment and correction compensation in a carton printing line is provided. Step S4 also includes tension measurement redundancy cross-calibration: the tension estimate obtained by the servo motor torque soft measurement is cross-calibrated with the measurement value of the tension sensor. When the deviation between the two exceeds 10% and continues for more than a preset time, an alarm is triggered and control is switched to the tension estimate. The edge detection delay is calculated in real time according to τ=Ls / v and discretized according to the control cycle, where Ls is the distance from the correction point to the edge sensor and v is the real-time linear velocity. Since v changes continuously during acceleration and deceleration, the delay parameter is refreshed frame by frame and rounded to the nearest control cycle number to ensure that the internal model delay of the Smith predictor is always synchronized with the actual delay.
[0020] Furthermore, a method for adaptive tension adjustment and correction compensation in a carton printing line is provided. In step S5, the feedback control quantity is generated by an incremental PID algorithm. During operation, the PID parameters are finely adjusted online according to the tension error and its rate of change based on fuzzy rules. Amplitude limiting and output rate of change limit are applied to the tension control quantity, and anti-saturation processing is implemented on the integral term. Step S5 further includes disturbance observation compensation: the tension measurement value is processed through the inverse model of the controlled object in the tension section to obtain the inverse model output, the difference between the inverse model output and the tension control quantity is processed through low-pass filtering to obtain the disturbance estimate (the low-pass filtering is used to suppress the measurement noise amplified by the inverse model differentiation operation), and the disturbance estimate is used as a compensation quantity in the synthesis to offset the load disturbance and friction disturbance; Step S5 also includes eccentricity compensation: the roller rotation frequency is calculated according to the real-time rotation speed of each traction roller, and the center frequency of the digital notch filter follows the rotation frequency in real time to suppress the periodic tension fluctuations of the rotation frequency and its harmonics caused by roller eccentricity. Since the rotation frequency changes continuously with the vehicle speed, the fixed frequency notch filter cannot cover it. The real-time tracking center frequency makes the notch filter always aligned with the eccentricity disturbance frequency during the speed change process, and a notch filter can be connected in parallel at the harmonic to suppress the secondary component.
[0021] Furthermore, a method for adaptive tension adjustment and correction compensation in a carton printing line is provided. In step S6, the correction deviation, after being estimated and compensated by Smith, and the correction feedback control together generate a correction command, which is output to the correction actuator. The correction command is divided into the following cases: When the correction actuator is a servo correction roller set in the continuous conveying section, the correction command is a swing angle command, the swing angle is limited to ±2° and the rate of change is limited to no more than 1° / s, to prevent excessive wrap angle changes from causing sudden tension changes, paperboard slippage or even paper breakage. When the correction actuator is the left and right paper feed rollers of a single sheet conveying section, the correction command is the differential speed command or differential pressure command of the left and right paper feed rollers; when the correction actuator is the left and right paper feed rollers, it also includes a skew correction step: the edge position is detected at two points before and after the paperboard conveying direction, the skew angle is obtained by the ratio of the difference between the two edge positions to the distance between the two detection points, the skew angle is converted into the additional differential speed of the left and right paper feed rollers, and then output after being superimposed with the deviation correction amount, so as to correct the deviation and the skew of the paperboard at the same time, and ensure the accuracy of subsequent printing registration and die-cutting.
[0022] Furthermore, a method for adaptive tension adjustment and correction compensation in a carton printing line is provided. In step S7, the tension disturbance caused by the correction action is estimated based on the equivalent tensile stiffness. The tension disturbance caused by the correction action is calculated as ΔTθ=EA·ΔL(θ) / L, where ΔL(θ)=2·Lt·sin²(θ / 2), where θ is the swing angle of the correction roller, Lt is the front and rear wrapping length of the correction roller, L is the segment length, and EA is the equivalent tensile stiffness. This tension disturbance is then superimposed on the tension setpoint. When the correction is expected to increase (decrease) the tension, the tension setpoint is pre-decreased (increased) accordingly to protect the tension ring from the correction impact. The lateral offset caused by the tension change is calculated as Δy=β·ΔT The calculation is performed, where β is the pre-calibrated tension-offset coupling coefficient, ΔT is the tension change, and the lateral offset is superimposed on the input of the correction loop; it also includes priority arbitration: when the correction action amplitude exceeds the first preset threshold, the integral action of the tension controller is paused for a preset duration; when the linear acceleration exceeds the second preset threshold, the rate of change limit of the correction command is tightened; and the closed-loop bandwidths of the tension loop and the correction loop are staggered by no less than 3 times to avoid mutual excitation and resonant coupling caused by the overlap of the two loop frequency bands.
[0023] Example 2 Taking a four-color printing, slotting, and die-cutting production line as an example: the width is 1.6m, the speed is 30~200m / min, there are 5 traction points along the line, forming 4 tension sections, each about 2m long. The control unit is an industrial computer with an EtherCAT bus, and the control cycle is 2ms. Each section is equipped with a tension sensor, each traction roller is equipped with an encoder, and an ultrasonic edge sensor is installed 0.8m downstream of the correction actuator. The continuous conveying section is equipped with a servo correction roller.
[0024] S1. Parameter Initialization: Establish a cardboard parameter library. Each record includes flute type, basis weight, moisture content, and corresponding historical identification parameters (equivalent tensile stiffness, object static gain, and converted inertia) and historical controller parameters. The current order is BC flute, basis weight 800g / m², and moisture content 10%. Search by "same flute type, basis weight difference less than 10%, moisture content difference less than 2%", find the BC flute historical record, and take its equivalent tensile stiffness 3.1×10. 4 N is used as the initial value; if the target is not hit, it is estimated based on the weight, width and model coefficient. For every 1% increase in moisture content, the initial value is reduced by 4%. The tension setting value for this order is calculated by multiplying the tension per unit width of 75 N / m by the width of 1.6 m to get 120 N. S2. Online Identification: During the paper feeding stage, the system operates stably at 30 m / min. An M-sequence excitation with an amplitude of ±1.5% is superimposed on the speed difference between adjacent traction rollers for 20 seconds (halved when hitting the database record). A recursive least squares method with a forgetting factor of 0.98 is used for online identification according to the discrete model described in claim 3. Convergence is determined when the parameter increment is less than 1% for 100 consecutive updates. The parameters are then locked and the excitation is canceled. The equivalent tensile stiffness of this order is calculated to be approximately 3.0 × 10⁻⁶. 4 N (deviation from initial value of 1.6%, mutually corroborating), the object's static gain is approximately 5 × 10⁻⁶. 4 Then, a ramp acceleration of 30→60m / min was performed. The equivalent moment of inertia was approximately 96kg, obtained from the ratio of the component in phase with the acceleration in the tension response (approximately 16N) to the acceleration (0.167m / s²). S3. Controller self-tuning: The desired closed-loop damping is 0.8, and the bandwidth is adjusted according to vehicle speed (8 rad / s for low speed, 12 rad / s for medium speed, and 15 rad / s for high speed). PI parameters are calculated based on pole configuration. For this order, the medium speed range yields Kp≈3.8×10⁻ 4 Ki≈2.8×10⁻³; Output limiting is ±10% of the master speed, and integral clamping is 90% of the limiting. S4. Operational data acquisition: The tension signal is filtered by a 50Hz power frequency notch and a 20Hz low-pass filter, and the edge signal is filtered by a 3-point median filter. The edge detection delay is calculated by dividing the sensor distance of 0.8m by the real-time linear velocity: approximately 0.32s at 150m / min and approximately 0.24s at 200m / min. The delay varies significantly with the vehicle speed, so it is refreshed every cycle. S5. Tension adaptive adjustment: Each cycle, three parts are combined into a tension control quantity: (1) Feedback control quantity, generated by incremental PID according to tension error. When the error is large and changes rapidly, Kp is increased and Ki is decreased. When the error is small, Ki is increased to eliminate steady error. The output is limited, integral anti-saturation and rate of change are limited; (2) Acceleration inertia feedforward, which is equal to the product of the converted inertia and the real-time acceleration. For example, when the acceleration is 0.5m / s², it can synchronously offset about 48N tension impact (up to 40% of the set value) without waiting for the feedback link to passively suppress; (3) Cascade speed difference feedforward. When the speed increases or decreases, each section is given synchronously with the master command according to its own steady-state speed difference bias. This bias is refreshed with the identification result of the equivalent tensile stiffness. The combined result is limited and then used as the speed difference command output of each traction roller; S6. Correction Compensation: The inner mold of the Smith predictor is a correction roller link with inertia. Its delay parameter is refreshed every cycle according to the edge detection delay obtained in S4, so as to achieve consistent delay compensation across the entire speed range. The correction deviation after Smith compensation is fed back by PI to generate a swing angle command, with a limit of ±2° and a change rate of no more than 1° / s, to drive the correction roller. S7. Decoupling and Coordination: On the one hand, the tension disturbance caused by the correction is calculated based on the equivalent tensile stiffness and swing angle. When the swing angle is 2°, it is about 5.6N (4.7% of the set value). It is then superimposed on the tension setpoint in real time to cancel it out. On the other hand, the lateral offset caused by the tension change is calculated based on the pre-calibrated coupling coefficient (0.025mm / N). When the tension change is 10N, it is about 0.25mm. It is then superimposed on the correction loop input in real time to cancel it out. The closed-loop bandwidths of the two loops are staggered by 4 times (tension loop 12rad / s, correction loop 3rad / s) to avoid mutual excitation.
Claims
1. A method for tension self-adaptive adjustment and deviation compensation of a carton printing line, characterized in that, Includes the following steps: S1. Parameter initialization: Establish a paperboard parameter library, which records paperboard feature information and corresponding historical identification parameters and historical controller parameters; obtain the paperboard feature information of the current order, search for matching records in the paperboard parameter library, and use the historical identification parameters in the matching record as the initial value when a match is found, and estimate the initial value of the equivalent tensile stiffness of the current paperboard when no match is found. S2. Online identification: During the paper feeding stage, an excitation signal is superimposed on the speed difference between adjacent traction rollers, the tension response is collected, and the equivalent tensile stiffness and static gain of the current paperboard are identified online using the recursive least squares method with a forgetting factor. The moment of inertia is also identified based on the tension response under the known acceleration command. S3. Controller self-tuning: Based on the equivalent tensile stiffness, the static gain of the object, and the reduced inertia, the parameters of the tension controller are self-tuned according to the desired closed-loop characteristics; S4. Operational Data Acquisition: Real-time acquisition of tension in each section, linear speed of each traction roller, edge position of the cardboard, and registration mark; and calculation of edge detection delay that varies with vehicle speed based on real-time linear speed. S5. Tension Adaptive Adjustment: The acceleration inertia feedforward, the cascade speed difference feedforward, and the feedback control quantity generated by the tension controller are combined into a tension control quantity, and the speed difference command of each traction roller is output; wherein, the acceleration inertia feedforward is determined by the converted inertia and the real-time acceleration, and the coefficient of the cascade speed difference feedforward is refreshed according to the online identification result of the equivalent tensile stiffness. S6. Correction compensation: Construct a Smith predictor. The internal model delay parameter of the Smith predictor is refreshed online with the vehicle speed according to the edge detection delay. The correction deviation after Smith prediction compensation and the correction feedback control together generate a correction command, which is output to the correction execution mechanism. S7. Decoupling and Coordination: Estimate the tension disturbance caused by the correction action based on the equivalent tensile stiffness, and feed the tension disturbance forward to the tension adjustment channel to cancel it out; estimate the lateral offset caused by the tension change based on the pre-calibrated tension-offset coupling coefficient, and feed the lateral offset forward to the correction channel to cancel it out, thereby achieving decoupling between the tension loop and the correction loop.
2. The method for adaptive tension adjustment and correction compensation of a carton printing line as described in claim 1, characterized in that, In step S1, the cardboard feature information of the current order is obtained, and a matching record is retrieved from the cardboard parameter library, as follows: The cardboard feature information includes flute type, basis weight, width, and moisture content; the retrieval is a fuzzy retrieval, and the matching conditions are the same flute type, basis weight difference less than 10%, and moisture content difference less than 2%. When a matching record is hit, the duration of the excitation signal in step S2 is halved; when no match is found, the initial value of the equivalent tensile stiffness is estimated according to the correction coefficients of flute type, face paper basis weight and liner paper basis weight, width, and moisture content, wherein for every 1% increase in moisture content, the initial value of the equivalent tensile stiffness is reduced by 3% to 5%.
3. The method for adaptive tension adjustment and correction compensation of a carton printing line as described in claim 1, characterized in that, In step S2, during the paper feeding stage, an excitation signal is superimposed on the speed difference between adjacent traction rollers, and the tension response is collected. The equivalent tensile stiffness and static gain of the current paperboard are identified online using a recursive least squares method with a forgetting factor, as detailed below: The excitation signal is a pseudo-random binary sequence and a bidirectional step signal superimposed on the speed difference between adjacent traction rollers, with an amplitude of 1% to 2% of the current linear velocity; The discrete model used for identification is T(k) = a·T(k-1) + b·δv(k-1-d) + c, where T(k) is the tension of the k-th beat, δv(·) is the speed difference, d is the delay beat number, and c is the steady-state bias. The forgetting factor of the recursive least squares method with forgetting factor is 0.95~0.
99. When the relative change of parameter increment is less than 1% for 100 consecutive beats, the identification is determined to be converged, the parameters are locked, and the excitation signal is canceled. After the identification is converged, the equivalent tensile stiffness and the static gain of the object are converted according to EA = b·L / Ts and K_T = b / (1-a), where L is the length of the section between adjacent traction rollers and Ts is the control cycle. The identification method of the converted inertia is as follows: execute the acceleration command with known slope, extract the component in phase with the acceleration in the tension response, and use the ratio of this component to the acceleration command as the equivalent inertia converted to the roller surface.
4. The method for adaptive tension adjustment and correction compensation of a carton printing line as described in claim 1, characterized in that, In step S3, the desired closed-loop characteristic is a second-order characteristic with a damping ratio of 0.7 to 0.
9. The closed-loop bandwidth is determined according to the constraint of the edge detection delay and is graded according to the vehicle speed. The tension controller parameters are obtained by pole configuration, including Kp=ωc / K_T and Ki=ωc² / (2·K_T), where ωc is the closed-loop bandwidth and K_T is the static gain of the object.
5. The method for adaptive tension adjustment and correction compensation of a carton printing line as described in claim 1, characterized in that, Step S4 also includes tension measurement redundancy cross-calibration: the tension estimate obtained by the servo motor torque soft measurement is cross-calibrated with the measurement value of the tension sensor. When the deviation between the two exceeds 10% and continues for more than a preset time, an alarm is triggered and control is switched to the tension estimate. The edge detection delay is calculated in real time according to τ=Ls / v and discretized according to the control cycle, where Ls is the distance from the correction point to the edge sensor and v is the real-time linear velocity.
6. The method for adaptive tension adjustment and correction compensation of a carton printing line as described in claim 1, characterized in that, In step S5, the feedback control quantity is generated by an incremental PID algorithm. During operation, the PID parameters are finely adjusted online according to the tension error and its rate of change based on fuzzy rules. Amplitude limiting and output rate of change limit are applied to the tension control quantity, and anti-saturation processing is implemented on the integral term. Step S5 further includes disturbance observation compensation: the tension measurement value is processed through the inverse model of the controlled object in the tension section to obtain the inverse model output, the difference between the inverse model output and the tension control quantity is filtered by low-pass to obtain the disturbance estimate, and the disturbance estimate is used as a compensation quantity to participate in the synthesis to offset the load disturbance and friction disturbance. Step S5 also includes eccentricity compensation: the roller rotation frequency is calculated according to the real-time rotation speed of each traction roller, and the center frequency of the digital notch filter follows the rotation frequency in real time to suppress the periodic tension fluctuations of the rotation frequency and its harmonics caused by roller eccentricity.
7. The method for adaptive tension adjustment and correction compensation of a carton printing line as described in claim 1, characterized in that, In step S6, the correction deviation, after being estimated and compensated by Smith, and the correction feedback control together generate a correction command, which is output to the correction actuator. The correction command is divided into the following cases: When the correction actuator is a servo correction roller installed in the continuous conveying section, the correction command is a swing angle command, the swing angle is limited to ±2° and the rate of change is limited to no more than 1° / s; When the correction actuator is the left and right paper feed rollers of a single sheet conveying section, the correction command is the differential speed command or differential pressure command of the left and right paper feed rollers; when the correction actuator is the left and right paper feed rollers, it also includes a skew correction step: the edge position is detected at two points before and after the paperboard conveying direction, the skew angle is obtained by the ratio of the difference between the two edge positions to the distance between the two detection points, the skew angle is converted into the additional differential speed of the left and right paper feed rollers, and then output after being superimposed with the deviation correction amount.
8. The method for adaptive tension adjustment and correction compensation of a carton printing line as described in claim 1, characterized in that, In step S7, the tension disturbance caused by the correction action is estimated based on the equivalent tensile stiffness. The tension disturbance caused by the correction action is calculated as ΔTθ=EA·ΔL(θ) / L, where ΔL(θ)=2·Lt·sin²(θ / 2), where θ is the swing angle of the correction roller, Lt is the front and rear wrapping length of the correction roller, L is the segment length, and EA is the equivalent tensile stiffness. This tension disturbance is superimposed on the tension setpoint. The lateral offset caused by the tension change is calculated as Δy=β·ΔT, where β is the pre-calibrated tension-offset coupling coefficient, and ΔT is the tension change. This lateral offset is superimposed on the input of the correction loop. The process also includes priority arbitration: when the correction action amplitude exceeds the first preset threshold, the integral action of the tension controller is paused for a preset duration; when the linear acceleration exceeds the second preset threshold, the rate of change limit of the correction command is tightened; and the closed-loop bandwidths of the tension loop and the correction loop are offset by at least 3 times.
