High-speed rewinding machine transmission-tension-vibration coordinated speed-up control method and system

Abstract

The application discloses a high-speed rewinding machine transmission-tension-vibration collaborative speed-up control method and system, and belongs to the technical field of automatic control of papermaking equipment. The method comprises the following steps: synchronously collecting multiple source signals; constructing a dynamic model containing transmission-tension, tension-vibration and vibration-transmission coupling coefficients, and performing online identification every 100 ms; performing multi-objective rolling optimization by using explicit model predictive control, and the cost function contains tension deviation, vibration amplitude and speed tracking penalty; after the effective value of vibration speed continuously exceeds the threshold, the vibration suppression weight is dynamically lifted by spectrum analysis to realize active damping; the speed is changed according to the diameter of the parent roll, the speed tracking is emphasized on the side of full roll and full speed, and the tail roll is dynamically decelerated; a double-time-scale closed loop of 20 ms control and 100 ms identification is formed. The system comprises a sensing layer, a preprocessing layer, an eMPC and RLS hybrid control layer, an execution layer and a real-time communication layer. The application realizes stable speed-up from 2500 m / min to 2600 m / min without changing the main structure, the paper breakage rate is reduced by 65.2%, and the vibration speed is reduced to below 2.8 mm / s.

Description

Technical Field

[0001] This invention relates to the field of automatic control technology for papermaking equipment, specifically to a method and system for coordinated speed-up control of transmission, tension, and vibration in high-speed rewinders. Background Technology

[0002] A search revealed that existing automatic control technologies for papermaking equipment have shortcomings in the following areas:

[0003] The lack of a dimension in multivariable coupled control: In the field of rewinder control, the "TNT" two-parameter coupling relationship between torque, pressure zone, and paper web tension has been academically studied, and some literature has also explored tension optimization strategies based on model predictive control. However, none of the above-mentioned existing technologies have incorporated "vibration" as an independent and strongly coupled dimension into the unified control framework. Specifically, existing solutions fail to explicitly model the dynamic disturbance of instantaneous fluctuations in the transmission system on the paper web tension, fail to quantify the alternating excitation effect of tension fluctuations transmitted through the roller system to the bearing housing / frame, and fail to consider the inner closed-loop effect of vibration state inversely affecting the transmission control accuracy through encoder signal noise, mechanical resonance, and other means. Therefore, existing technologies are essentially "dimensional reduction control," sacrificing real-time active control of the vibration dimension in exchange for single-point optimization of tension or speed. At high-speed conditions of ≥2500m / min, this lack of dimension leads to a sharp deterioration in control quality.

[0004] The limitations of the "post-event response" paradigm in vibration control: Most existing vibration control schemes employ an open-loop triggering mode of "detection → judgment → deceleration," which is a typical passive response. While some schemes propose the concept of "advance prediction," their essence remains a logic-based early warning system with fixed thresholds. This means issuing a warning signal or executing a deceleration command when a characteristic value exceeds a limit. Their prediction logic does not involve active intervention in the evolution trend of the vibration spectrum, and they cannot achieve proactive suppression by adjusting the system's damping characteristics before vibration energy accumulates. There remains a significant technological leap between "predictive alarm" and "active suppression."

[0005] The multi-objective extension of Model Predictive Control (MPC) is insufficient: Although some studies have introduced MPC into the tension control of rewinding machines, their optimization objectives are focused on minimizing tension fluctuations. While a few schemes mention multi-objective optimization, none have incorporated the effective values ​​of vibration velocity / acceleration as real-time optimization objectives into the cost function. More importantly, standard MPC requires solving a quadratic programming problem online, whose computational load far exceeds the real-time computing power limit of industrial programmable controllers. This has resulted in this technical approach remaining largely at the academic simulation stage, with no precedent for engineering deployment.

[0006] The economic pain point of speed-up upgrades: Existing solutions for increasing the speed of rewinding machines from 2500m / min to 2600m / min mostly rely on replacing the drive motor with a high-power one, reinforcing the mechanical roller system, or upgrading the hydraulic system, resulting in extremely high costs per unit.

[0007] Therefore, the industry urgently needs a low-cost solution that focuses on algorithm upgrades, keeps the main mechanical structure unchanged, and moderately enhances the sensing layer. Summary of the Invention

[0008] The purpose of this invention is to solve the bottleneck problem of speed increase caused by the strong coupling of transmission fluctuations, tension instability, and vibration divergence under high-speed operating conditions. By constructing a multidimensional coupled dynamic model, designing a hybrid architecture of explicit model predictive control (eMPC) and recursive least squares (RLS), and introducing a vibration advance active suppression mechanism based on spectral characteristics, a stable speed increase to 2600m / min can be achieved without changing the main mechanical structure of the rewinder, significantly reducing the paper breakage rate and equipment degradation rate.

[0009] To achieve the above objectives, the present invention provides the following technical solution:

[0010] A high-speed rewinder transmission-tension-vibration coordinated speed-up control method includes the following steps:

[0011] S1. Collect multi-source data on the operating status of the rewinding machine through a distributed sensor network, and then input the data into the central controller after clock synchronization.

[0012] The multi-source data includes incremental encoder pulse signals, tension sensor analog signals, accelerometer vibration signals, and laser rangefinder distance signals; wherein the encoder resolution is 2048 pulses / revolution, and the tension sensor accuracy is... The frequency response range of the accelerometer is Clock synchronization adopts the IEEE 1588 precision time protocol. Control-related channels acquire data at a sampling frequency of at least 1 kHz, and acceleration channels at a sampling frequency of at least 10 kHz, covering the 0.5 Hz to 5 kHz frequency band. The data preprocessing layer performs local high-speed sampling and buffering of the acceleration channel, and performs FFT at a 100 ms analysis cycle to extract the main frequency and amplitude characteristics within the 0.5 Hz to 5 kHz frequency band before uploading to the central controller. Control data, after extraction and filtering, is imported into the central controller at a rate of at least 1 kHz. This high-speed data import uses a Profinet or EtherCAT real-time industrial Ethernet bus, with a communication cycle of no more than 1 ms. Data preprocessing includes signal conditioning, synchronization, filtering, acceleration integration to obtain vibration velocity, and calculation of the effective value of vibration velocity in a 20 ms window. .

[0013] S2. Construct a three-state coupled dynamic model that includes transmission state variables, tension state variables, and vibration state variables. The model includes at least the disturbance coefficient of transmission fluctuation on tension, the excitation coefficient of tension fluctuation on vibration, and the influence coefficient of vibration on transmission accuracy. The model parameters are identified and updated online at a fixed time interval of 100ms.

[0014] The three-state coupled dynamics model adopts a state-space form: ;

[0015] Where, the state variable x = [V, T, v] rms ] T ;

[0016] V is the linear velocity of the paper web. For paper web tension, This represents the effective value of the bearing housing vibration velocity.

[0017] Control input ;

[0018] This is the motor torque command. For hydraulic proportional valve pressure command; disturbance This includes paper roll eccentricity and paper web basis weight fluctuations. Angular velocity. From linear velocity and the remaining diameter of the mother roll The conversion is as follows: ,in The unit is m / min. The unit is rad / s. The unit is mm. Remaining diameter of the master roll. It is measured in real time by a laser rangefinder and serves as a measurable slowly varying parameter or disturbance input.

[0019] System Matrix The explicit inclusion of three types of coupling coefficients. Let... Indicates the first The state variable is the first... The influence of the derivatives of the state variables is as follows:

[0020] The disturbance coefficient of transmission fluctuation on tension, characterizing the tension change caused by linear velocity fluctuation;

[0021] The excitation coefficient of tension fluctuation on vibration, characterizing the excitation generated by the alternating tension component on the bearing housing through the roller system;

[0022] The vibration influence coefficient on transmission accuracy represents the effect of vibration on transmission control accuracy through encoder noise, mechanical resonance, and other means.

[0023] The aforementioned coupling coefficients were identified online using the Recursive Least Squares (RLS) algorithm, with an identification period of 100ms. The RLS parameter update formula is as follows:

[0024] ;

[0025] ;

[0026] ;

[0027] in, Forgetting factor, take Effective value of vibration velocity A first-order envelope dynamic approximation is used to describe the following:

[0028] ;

[0029] in, The vibration envelope time constant is The excitation gain of the vibration due to tension fluctuation. This is the damping control gain. The equation shows... The change depends on the accumulation of tension fluctuation energy and the damping control effect, thus giving it a clear physical meaning as a state variable.

[0030] S3. Multi-objective collaborative rolling optimization based on explicit model predictive control (eMPC). The coupling coefficient and cost function weights are used as adjustable parameters. Offline, the model parameters are discretized according to their range and weight levels. For each combination, a parameterized piecewise affine function library is solved. During the online control cycle, based on the range of model parameters identified by the current RLS and the current weight level, the corresponding piecewise affine function is selected, and control commands are obtained through table lookup and interpolation. The basic control cycle is 20ms, and the prediction time domain is... One control cycle, with a control time domain of [missing information]. Each control cycle has a cost function that includes at least a tension deviation penalty term, a vibration amplitude penalty term, and a velocity tracking deviation penalty term.

[0031] The cost function of eMPC is:

[0032] ;

[0033] in, To predict the time domain, To control the time domain; Here are the weighting coefficients, where Dynamically lifts when active vibration suppression is triggered; For the rewinder linear speed, This is a reference value for linear velocity.

[0034] The constraints include:

[0035] Motor torque limiting: ;

[0036] Limit on the rate of change of linear velocity: ;

[0037] Tension safety boundary: .

[0038] When solving offline, the above optimization problem is transformed into a multi-parameter quadratic programming problem. The model parameter intervals are arranged according to the coupling coefficient. The typical range of variation is divided into several discrete intervals, and the weighting levels are based on... The combination is divided into several levels, such as vibration suppression weight. It is divided into 5 levels: 1, 2, 3, 4, and 5 times. For each combination of model parameter range and weight level, the corresponding piecewise affine function is obtained:

[0039] ;

[0040] in, For the range index of model parameters, This is the index for the weight level. During the 20ms online control cycle, it is based on the current state. The RLS-updated model parameters belong to the specified interval and current weight level. The corresponding piecewise affine function is selected, the effective region is found, and control commands are calculated using table lookup and linear interpolation. This eliminates the need for online quadratic programming, enabling real-time control within the computational constraints of industrial PLCs. It assumes that model parameters remain constant in the prediction time domain, and reselects the affine function based on the latest identification results in each control cycle, achieving rolling optimization.

[0041] S4. Perform proactive vibration suppression: Set the effective value of vibration velocity. The normal operating baseline value is Trigger threshold is The inherent frequency is pre-stored in the operating condition knowledge storage module, or obtained through offline modal testing and online spectrum peak tracking.

[0042] Determine the acceleration spectrum characteristics uploaded from the data preprocessing layer: when The trigger threshold has not yet been exceeded, but the deviation between the main frequency and the inherent frequency of the roller system or frame is within... If the dominant frequency amplitude shows an upward trend within a range of 100ms FFT analysis periods, it is determined to be a vibration energy accumulation trend, and the vibration suppression weight in the cost function is adjusted accordingly. The speed is increased from the initial setting (1x) to the first increased setting (2x). If the detected value exceeds the trigger threshold for three consecutive 20ms sampling windows, the acquired spectrum analysis results are used again. If the dominant frequency amplitude exceeds twice the normal operating condition baseline amplitude, a resonance risk is confirmed, and the vibration suppression weight is adjusted. The gear shifts from the current gear to the second shift gear (any gear between 3 and 5x), without exceeding the preset maximum gear; if no pre-triggered shift is performed, the gear shifts directly from the initial gear (1x) to the second shift gear. Vibration suppression weight. It is divided into 5 levels: 1, 2, 3, 4, and 5 times, with the maximum increase being 5 times the initial value.

[0043] The dynamic lifting is achieved by selecting a pre-generated piecewise affine function corresponding to the weight level. The eMPC updates the control commands based on the new weight level through table lookup and interpolation, adjusting the pressure in the pressure zone or the damping cylinder via a hydraulic proportional valve to change the system's damping characteristics and implement active damping regulation, thus suppressing vibration energy before it accumulates. The sampling window has a 20ms control cycle, i.e., three consecutive 20ms windows, totaling 60ms.

[0044] S5. Implement a segmented speed-changing strategy based on the remaining diameter of the master roll. The remaining diameter of the master roll is obtained using a laser rangefinder. .

[0045] when When the speed reaches 2600 m / min at full speed, the cost function is... Increased, with a focus on speed tracking;

[0046] when At that time, maintain full speed. Increase, focusing on tension stability;

[0047] when When the paper type and working conditions are met, a dynamic deceleration curve is activated, and the speed is reduced to the roll change speed according to a preset slope. The roll change speed is set to 600-800 m / min, and the deceleration slope is set to 30 (m / min) / s. These values ​​can be preset in the working condition knowledge storage module according to the paper type and working conditions.

[0048] S6. The high-speed control calculation of S3 is executed with a second control cycle of 20ms, and the online identification of model parameters of S2 is executed with a first control cycle of 100ms, forming a closed-loop iteration with dual time scales.

[0049] Dual-time-scale closed loop: eMPC lookup interpolation is performed to output control commands during the second control cycle of 20ms; online identification of RLS model parameters is performed during the first control cycle of 100ms to update the system matrix. The coupling coefficients in the model are used to select the corresponding model parameter range in the parameterized piecewise affine function library, forming an iterative closed loop of "fast control-slow identification".

[0050] A high-speed rewinder transmission-tension-vibration coordinated speed-up control system, used to execute the above method, includes:

[0051] The sensing layer is equipped with a high-resolution incremental encoder, a high-precision tension sensor, a wideband piezoelectric accelerometer, and a laser rangefinder.

[0052] The data preprocessing layer includes a hardware clock synchronization module that supports the IEEE 1588 precise time protocol. This module is used to condition, synchronize, and filter the sensor signals, and to perform high-speed sampling, buffering, and local FFT on the acceleration channel to extract the main frequency and amplitude features.

[0053] The central control layer, equipped with a hybrid architecture of eMPC and RLS, is used to execute the method steps S2 to S6 and receive the data processed by the sensing layer and the data preprocessing layer after S1. The central control layer includes a working condition knowledge storage module, which is used to store the complete working condition data after each roll change, the inherent frequency of the roll system or frame, the weight level, and the offline optimized eMPC parameterized piecewise affine function library.

[0054] The execution layer includes servo drives for each roller and hydraulic proportional valves;

[0055] The real-time communication layer uses the Profinet or EtherCAT protocol with a communication cycle of no more than 1ms, connecting the sensing layer, data preprocessing layer, central control layer, and execution layer.

[0056] This invention is the first to incorporate "vibration" as an independent and strongly coupled state variable into the unified control framework of a rewinding machine, constructing a three-state coupled dynamic model of transmission, tension, and vibration, and explicitly modeling the bidirectional interaction effect among the three. It breaks through the traditional passive vibration suppression paradigm of "post-event deceleration" and pioneers a vibration proactive suppression mechanism based on spectral evolution trends. Dynamic weight adjustment is triggered by the dual conditions of "effective value exceeding the limit + duration," achieving predictive damping regulation of vibration energy. Simultaneously, a hybrid architecture of explicit model predictive control (eMPC) and recursive least squares (RLS) is introduced, solving the multi-objective optimization problem offline as a parameterized piecewise affine function library. Online, only table lookup and interpolation are required, achieving coordinated rolling optimization of tension, speed, and vibration under the constraints of industrial PLC computing power. Finally, with zero modification to the main mechanical structure and moderate enhancement of the sensing layer, a traditional modification scheme is achieved. The cost-effectiveness of the rewinding machine was reduced, and the operating speed was steadily increased from 2500m / min to 2600m / min. At the same time, the tension accuracy, vibration suppression effect and adaptability to multiple paper types were significantly improved. The system systematically solved the bottleneck problem of speed improvement caused by strong coupling of multiple variables under high-speed conditions.

[0057] Compared with the prior art, the beneficial effects of the present invention are:

[0058] 1. Under the condition that the main mechanical structure remains completely unchanged, the present invention achieves an increase in speed from 2500m / min to 2600m / min, with a net increase in daily production capacity of about 4% and an annual increase in production of about 12,200 tons;

[0059] 2. The fixed costs of this invention are diluted due to increased production volume, and the labor and depreciation costs per ton of paper are reduced simultaneously; the number of paper breaks is reduced by 65.2%, and raw material losses are significantly reduced; vibration is proactively suppressed to extend the life of key vulnerable parts and reduce spare parts maintenance costs;

[0060] 3. The tension fluctuation amplitude of the equipment of the present invention has been reduced from ±3.5% to ±1.8%, the end face deviation of the finished roll has been improved from ≤2.0mm to ≤1.2mm, and the average number of paper breaks per day has been sharply reduced from 2.3 times to 0.8 times;

[0061] 4. This invention breaks through the traditional paradigm of "threshold triggering - passive speed reduction" and pioneers a new path of spectrum prediction, dual-condition triggering and active damping adjustment. The effective value of the bearing housing vibration velocity is stabilized from 4.2 mm / s to below 2.8 mm / s, a reduction of 33.3%.

[0062] 5. The main mechanical structure of the rewinder of this invention requires no modification, and the cost of a single unit modification is 10% to 15% of that of traditional mechanical replacement solutions. It is suitable for 30 to 200 g / m³. 2 It can adapt to different paper weights and can complete model adaptation in a very short time after changing rolls;

[0063] 6. This invention does not simply superimpose RLS identification, eMPC lookup, and vibration detection. Instead, it incorporates the effective vibration velocity value into the eMPC prediction model and cost function through a three-state coupled model, enabling eMPC to simultaneously predict tension deviation, velocity deviation, and vibration energy evolution in the prediction time domain. By pre-triggered dynamic weighting based on spectral trends, vibration suppression is transformed from "post-event deceleration" to "pre-event damping adjustment." Furthermore, through a parameterized piecewise affine function library, the above multi-objective optimization is solved offline and executed online via lookup, resolving the computational power constraints of industrial PLCs. The combination of these three elements produces a synergistic effect that cannot be achieved by any single technique: while maintaining a speed of 2600 m / min, the effective vibration value is reduced from 4.2 mm / s to below 2.8 mm / s, and tension fluctuations are reduced from... Down to The average number of paper breaks per day decreased from 2.3 times to 0.8 times. Detailed Implementation

[0064] The present invention will be further described in detail below with reference to embodiments, but the scope of protection of the present invention is not limited thereto.

[0065] Example 1: Speed-up verification of Valmet rewinding machine in a paper company

[0066] The target of the implementation is a Valmet high-speed rewinding machine, with an original maximum stable operating speed of 2500 m / min. Implementation steps: Install high-precision sensing equipment, deploy an eMPC+RLS hybrid controller, perform initial parameter self-tuning (run stably at 2500 m / min and collect 100 sets of operating condition data), and then gradually increase the speed to 2600 m / min.

[0067] The specific implementation process is as follows:

[0068] A 2048 pulse / revolution incremental encoder was added to the Valmet high-speed rewinder. Tension sensor, Accelerometer and laser rangefinder sensors are used. Clock synchronization is achieved via IEEE 1588, and data is transmitted to the central controller via EtherCAT bus with a communication cycle of no more than 1ms. The sampling rate of relevant control channels is no less than 1kHz, and the sampling rate of the acceleration channel is no less than 10kHz. The data preprocessing layer performs local FFT on the acceleration channel and uploads the main frequency and amplitude characteristics every 100ms. The central controller uses an eMPC+RLS hybrid architecture. It first runs stably at 2500m / min, collecting 100 sets of operating condition data. The initial coupling coefficient is identified using RLS, and an initial three-state coupled dynamic model is established. Then, the eMPC parameterized piecewise affine function library is solved offline to generate lookup table data. The speed is gradually increased to 2600m / min. During operation, eMPC lookup table interpolation outputs control commands every 20ms, and RLS parameters are updated every 100ms. When the effective value of vibration velocity exceeds [a certain value] for three consecutive 20ms windows... At this point, the vibration suppression weight is further increased to a second increase level based on the existing weight level, triggering active damping adjustment. Through the above technical solution, tension fluctuations are reduced from... Down to Vibration from Down to The number of paper breaks decreased from 2.3 times / day to 0.8 times / day.

[0069] The key performance indicators and actual performance comparison data are shown in Table 1 below (all data have been calibrated by field instruments):

[0070] Table 1. Comparison of key performance indicators before and after implementation of the high-speed rewinding machine.

[0071]

[0072] To verify The accuracy of the first-order envelope dynamic approximation model was compared and tested under typical working conditions. The results are shown in Table 2.

[0073] Table 2 Accuracy Comparison Test of First-Order Envelope Dynamic Approximation Model

[0074]

[0075] As can be seen from the table, the first-order envelope dynamic approximation performs well under typical operating conditions. The prediction errors are all less than 5%, proving that they have clear physical meaning and engineering applicability as state variables.

[0076] To verify the advantages of pre-triggered dynamic weight boosting over traditional threshold-triggered deceleration, a comparative test was conducted, and the results are shown in Table 3 below:

[0077] Table 3 Comparison Test of Different Vibration Suppression Strategies

[0078]

[0079] As shown in the table, under pre-triggering conditions, the vibration peak value decreased from 4.2 mm / s to 3.6 mm / s, indicating that the vibration effective value was raised in advance before exceeding the limit. This can effectively suppress the accumulation of vibration energy; at the same time, it avoids the production capacity loss caused by traditional speed reduction.

[0080] PWA function library generation and online lookup example: Model parameter ranges are... , , Divided into 3×3×3=27 intervals; weighted levels according to It is divided into 5 levels: 1, 2, 3, 4, and 5 times. Each category is divided into 3 levels, resulting in a total of 45 weight combinations. A multi-parameter quadratic programming problem is solved offline for each combination to obtain a piecewise affine function. During the online 20ms control cycle, it is only necessary to determine the results based on the RLS identification. Determine based on the spectrum trend prediction module The output can be obtained by looking up the table and performing linear interpolation. No need for online quadratic programming solutions.

[0081] Example 2: Real-world verification of adaptive switching between multiple paper types

[0082] For different paper weights, a 72-hour continuous operation comparison was conducted using both traditional PID control and the method of this invention. The method of this invention, after roll change, [achieved a result in approximately...]. Within 1 to 2 100ms identification cycles, the online identification and convergence of model parameters are completed, and there is no significant tension disturbance during the switching process.

[0083] The measured data are shown in Table 4:

[0084] Table 4 Comparison of Real-World Verification Data for Adaptive Switching of Multiple Paper Types (PID vs. This Invention, 72h Continuous Operation)

[0085]

[0086] As can be seen from the above embodiments, the method and system provided by the present invention significantly improve tension stability, vibration suppression effect and finished product qualification rate under high-speed operating conditions, and have good adaptability to different paper weights.

Claims

1. A high-speed rewinder transmission-tension-vibration coordinated speed-up control method, characterized in that, Includes the following steps: S1. Multi-source data on the operating status of the rewinder is collected through a distributed sensor network and then fed into the central controller after clock synchronization. The multi-source data includes at least the paper web linear speed signal, paper web tension signal, acceleration vibration signal and remaining diameter signal of the master roll. The acceleration channel is sampled and buffered locally at high speed, and then subjected to fast Fourier transform according to a set analysis period. The main frequency and amplitude features are extracted and then uploaded to the central controller. The control data is extracted and filtered and fed into the central controller at a rate of not less than 1kHz. S2. Construct a three-state coupled dynamic model that includes transmission state variables, tension state variables, and vibration state variables. The state variables are: x=[V,T,v rms ] T ; in, For the paper web linear velocity, For paper web tension, The effective value of the bearing housing vibration velocity; the model explicitly includes the disturbance coefficient of transmission fluctuation on tension. The excitation coefficient of tension fluctuation on vibration And the influence coefficient of vibration on transmission accuracy And trigger the online identification and update of model parameters in the first control cycle; S3. Multi-objective collaborative rolling optimization is performed based on explicit model predictive control (eMPC). The coupling coefficient and cost function weights are used as adjustable parameters. The model parameters are discretized offline according to the range of parameters and weight levels. Each combination is solved as a parameterized piecewise affine function library. In the online control cycle, the corresponding piecewise affine function is selected according to the range of parameters and weight level identified online. The control command is obtained by table lookup and interpolation. The cost function includes at least tension deviation penalty term, vibration amplitude penalty term, and velocity tracking deviation penalty term. The online control cycle is the second control cycle, and the second control cycle is shorter than the first control cycle. S4. Execute proactive vibration suppression: Acquire acceleration spectrum characteristics, and determine the vibration energy accumulation trend based on the deviation between the main frequency and the natural frequency of the roller system or frame, as well as the changing trend of the main frequency amplitude and / or frequency band energy; when the pre-trigger condition is met, before the effective value of the vibration velocity exceeds the trigger threshold, dynamically increase the vibration suppression weight in the cost function from the initial level to the first level; when the effective value of the vibration velocity exceeds the trigger threshold and continues for a preset sampling window, further dynamically increase the vibration suppression weight from the current level to the second level, without exceeding the preset maximum level; if no pre-trigger increase is performed, directly increase from the initial level to the second level; the dynamic increase is achieved by selecting a pre-generated piecewise affine function corresponding to the weight level, so as to change the system damping characteristics through the execution layer and implement active damping adjustment; S5. Implement a segmented speed change strategy based on the remaining diameter of the master roll: increase the weight of the speed tracking deviation penalty term during the full roll stage, maintain full speed and increase the weight of the tension deviation penalty term during the intermediate stage, and start the dynamic deceleration curve during the tail roll stage. S6. The high-speed control calculation of S3 is executed in the second control cycle, and the online identification of model parameters of S2 is executed in the first control cycle, forming a closed-loop iteration with dual time scales.

2. The method according to claim 1, characterized in that, In step S1, the relevant channel is controlled to acquire data at a sampling frequency of not less than 1 kHz, and the acceleration channel is controlled to acquire data at a sampling frequency of not less than 10 kHz to cover the 0.5 Hz to 5 kHz frequency band. The data preprocessing layer performs local high-speed sampling and caching of the acceleration channel, and performs FFT at a 100ms analysis cycle to extract the main frequency and amplitude features before uploading them to the central controller; the communication cycle of the real-time industrial Ethernet bus is no more than 1ms; the sampling window has a 20ms control cycle.

3. The method according to claim 1, characterized in that, In step S1, the multi-source data includes incremental encoder pulse signals, tension sensor analog signals, accelerometer vibration signals, and laser rangefinder distance signals; wherein the encoder resolution is 2048 pulses / revolution, and the tension sensor accuracy is... The frequency response range of the accelerometer is Clock synchronization adopts the IEEE 1588 precision time protocol and is integrated into the central controller via a Profinet or EtherCAT real-time industrial Ethernet bus.

4. The method according to claim 1, characterized in that, In step S2, the three-state coupled dynamic model adopts a state-space form: ； Where, control input u=[u m ,u p ] T ; u m This is the motor torque command. For hydraulic proportional valve pressure command; angular velocity From linear velocity and the remaining diameter of the mother roll The conversion is as follows: ,in The unit is m / min. The unit is rad / s. Unit is mm; remaining diameter of the master roll The laser rangefinder sensor measures the data in real time, serving as a measurable slowly varying parameter or disturbance input. The coupling coefficients are identified online using the Recursive Least Squares (RLS) algorithm, with an identification period of 100ms. The RLS parameter update formula is as follows: ； ； ； in, The forgetting factor is set at 0.95-0.

99. RMS value of vibration velocity A first-order envelope dynamic approximation is used to describe the following: ； in, The vibration envelope time constant is The excitation gain of the vibration due to tension fluctuation. This is the damping control gain.

5. The method according to claim 1, characterized in that: In step S3, the cost function of eMPC is: ； in, To predict the time domain, To control the time domain; Here are the weighting coefficients, where Dynamically lifts when active vibration suppression is triggered; This is a reference value for linear velocity; The constraints must include at least: motor torque limiting. Linear velocity change rate limit Tension safety boundary ; When solving offline, the optimization problem is transformed into a multi-parameter quadratic programming problem, and the model parameter intervals are arranged according to the coupling coefficient. The typical range of variation is divided into several discrete intervals, and the weighting levels are based on... The combinations are divided into several levels; for each combination of model parameter ranges and weight levels, the corresponding piecewise affine function is obtained by solving: u=F i,j x+G i,j ; in, For the range index of model parameters, For weighted level indexes; during the online control cycle, based on the current state The RLS-updated model parameters belong to the specified interval and current weight level. The corresponding piecewise affine function is selected, the effective region is found, and control commands are calculated using table lookup and linear interpolation. .

6. The method according to claim 1, characterized in that: In step S4, the pre-triggering conditions include: the deviation between the main frequency and the natural frequency of the roller system or frame is within... Within a certain range, and the main frequency amplitude shows an upward trend over multiple consecutive FFT analysis cycles, and the effective value of the vibration velocity has not exceeded the trigger threshold; The trigger threshold is set to The normal operating baseline value is When the detected value exceeds the trigger threshold for three consecutive sampling windows, the acquired spectrum analysis results are used again, and the vibration suppression weight is raised from the current level to the second level, without exceeding the preset maximum level; the sampling window has a 20ms control cycle. Vibration suppression weight It is divided into 5 levels: 1x, 2x, 3x, 4x, and 5x. The initial level is 1x, the first level is 2x, the second level is any level from 3x to 5x, and the maximum level is 5x. The inherent frequency is pre-stored in the operating condition knowledge storage module or obtained through offline modal testing and online spectrum peak tracking.

7. The method according to claim 1, characterized in that: In step S5, the remaining diameter of the mother roll At that time, the system runs at full speed and the weight of the speed tracking deviation penalty term is increased. The full speed is the target speed of 2600 m / min. At the same time, maintain full speed and increase the weight of the tension deviation penalty term; At that time, the dynamic deceleration curve is activated, and the speed is reduced to the roll change speed according to the preset slope; The roll changing speed is set to 600-800 m / min, and the deceleration slope is set to 30 (m / min) / s.

8. A high-speed rewinder transmission-tension-vibration coordinated control system, used to execute the method according to any one of claims 1 to 7, characterized in that, include: The sensing layer is equipped with an incremental encoder, a tension sensor, a broadband piezoelectric accelerometer, and a laser rangefinder. The data preprocessing layer includes a hardware clock synchronization module that supports the IEEE 1588 precise time protocol. This module is used to condition, synchronize, and filter the sensor signals, and to perform high-speed sampling, buffering, and local FFT on the acceleration channel to extract the main frequency and amplitude features. The central control layer includes an online least squares (RLS) identification module, an explicit model predictive control (eMPC) parametric lookup table interpolation module, a spectrum trend prediction module, a weight level selection module, and a working condition knowledge storage module. The RLS online identification module is configured to update the coupling coefficients in the three-state coupled dynamics model in the first control cycle. The spectrum trend prediction module is configured to determine the vibration energy accumulation trend based on the deviation between the dominant frequency and the natural frequency, as well as the amplitude and / or frequency band energy change trend of the dominant frequency. The weight level selection module is configured to select the corresponding weight level when the pre-triggering condition or the effective value exceeding the limit condition is met. The eMPC parametric lookup table interpolation module is configured to select piecewise affine functions based on the current model parameter range and weight level, and output control commands through lookup table and interpolation. The working condition knowledge storage module is used to store the working condition data after roll change, the natural frequency of the roll system or frame, the weight level, and the offline optimized eMPC parametric piecewise affine function library. The execution layer includes servo drives for each roller and hydraulic proportional valves; The real-time communication layer uses the Profinet or EtherCAT protocol with a communication cycle of no more than 1ms, connecting the sensing layer, data preprocessing layer, central control layer, and execution layer.

9. The system according to claim 8, characterized in that: The central control layer is configured to: execute eMPC lookup interpolation to output control commands with a second control cycle of 20ms, execute online identification of RLS model parameters with a first control cycle of 100ms, and select the corresponding model parameter interval in the parameterized piecewise affine function library accordingly, forming a dual-timescale iterative closed loop of fast control and slow identification.

10. The system according to claim 8, characterized in that: The data preprocessing layer performs FFT on the acceleration channel at a 100ms analysis period and uploads the main frequency and amplitude characteristics; the real-time communication layer uses Profinet or EtherCAT protocol, with a communication period of no more than 1ms; the encoder resolution in the sensing layer is 2048 pulses / revolution, and the tension sensor accuracy is... The frequency response range of the accelerometer is .

Images