An automatic leveling and shearing control system for medium-thick plate processing

Abstract

This invention discloses an automatic leveling and shearing control system for medium-thick plate processing, relating to the field of metal sheet processing technology. The control system includes a central controller that identifies the dynamic mechanical model of the plate online based on real-time data from a multimodal sensing unit. Based on the online identification model, within a limited prediction window, it uses a segmented model predictive control DMPC to calculate and output the optimal displacement and force sequences for each leveling roller unit. Simultaneously, it transmits predicted information on the plate surface morphology and stress distribution to the traction and shearing units for dynamic correction of shearing trigger timing and tool position compensation. This invention uses segmented model predictive leveling driven by multimodal short-time online identification to construct a physically correlated short-time dynamic model in real time, predicting the surface shape and stress for several future steps. It finely allocates control quantities for each roller to actively suppress residual curvature and stress unevenness, overcoming the failure of empirical or open-loop leveling under variations in material batches, temperature, and linear speed.

Description

Technical Field

[0001] This invention relates to the field of metal sheet processing technology, specifically to an automatic leveling and shearing control system for processing medium and heavy plates. Background Technology

[0002] In traditional production lines for medium and heavy plate processing, leveling and shearing are typically two relatively independent subsystems: leveling often employs segmented rigid rollers or hammer leveling, with control strategies primarily based on empirical PID or open-loop preset parameters; shearing relies mainly on length gauges and timed triggering, with insufficient compensation for material elastic rebound and instantaneous stress changes. Common problems include: unstable leveling with variations in line speed or material batches; residual bending and stress concentration affecting cutting accuracy; local defects such as welds or hard spots causing tool damage and high scrap rates; and system debugging relying heavily on manual experience and being time-consuming. Existing solutions lack closed-loop collaborative strategies based on multimodal short-term prediction, rapid event-driven compensation for sudden local defects, and online model identification and self-learning optimization capabilities. Therefore, it is difficult to simultaneously meet the requirements for flatness, shearing accuracy, and production capacity under conditions of high line speed, large thickness range, and mixed materials.

[0003] Patent CN105170712B discloses an automatic leveling and ranging shearing machine and its control method. The above patent realizes the full automation of leveling and has the advantages of high leveling accuracy, good consistency and high efficiency.

[0004] The aforementioned patent uses a fully automated leveling device and method with one-time and two-time leveling. By unifying the data from the two leveling operations, it achieves consistent leveling of the sheet material during shearing. The leveling efficiency is high, the accuracy is high, and the automation level is extremely reliable. However, it suffers from problems such as using segmented rigid rollers or hammer leveling for leveling, relying mainly on empirical PID or open-loop preset parameters for control strategies, and relying heavily on length gauges and timed triggering for shearing. It also suffers from insufficient compensation for material elastic rebound and instantaneous stress changes.

[0005] To address this, this application proposes an automatic leveling and shearing control system for medium-thick plate processing that overcomes the failure of empirical or open-loop leveling under variations in material batches, temperature, and linear speed. Summary of the Invention

[0006] The purpose of this invention is to provide an automatic leveling and shearing control system for medium and heavy plate processing, in order to solve the technical problems mentioned in the background art, such as unstable leveling when the linear speed or material batch changes, residual bending and stress concentration affecting cutting accuracy, local defects such as welds or hard spots causing tool damage and high scrap rate, and system debugging relying on manual experience and being time-consuming.

[0007] To achieve the above objectives, the present invention provides the following technical solution: an automatic leveling and shearing control system for processing medium-thick plates, the control system comprising:

[0008] A multimodal sensing unit is used to collect real-time data on the board's profile, thickness, temperature, and stress-strain along the length and width of the board.

[0009] The segmented controllable leveling roller unit consists of several independently driven roller sections with force and position feedback;

[0010] Local compensation module, used to provide instantaneous mechanical compensation for local protrusions, welds or hard spots;

[0011] The traction and shearing units are equipped with closed-loop control of position, speed, and force, while the shearing unit has closed-loop control of tool position and tool force.

[0012] The central controller communicates with each unit via a real-time bus and is configured to: identify the dynamic mechanical model of the board material online based on real-time data from the multimodal sensing unit; calculate and output the optimal displacement and force sequence of each leveling roller unit using a segmented model prediction control within a limited prediction window based on the online identification model; and transmit the predicted information of the board surface morphology and stress distribution to the traction and shearing units for dynamic correction of shearing trigger timing and blade position compensation, thereby realizing closed-loop coordinated control of leveling and shearing.

[0013] Preferably, the multimodal sensing unit includes: a laser profile scanner, a thickness measurement device, a distributed optical fiber or strain gauge sensor, an infrared temperature sensor, and a vision camera; the vision camera, in conjunction with a lightweight neural network model, is used to identify and locate welds, hard spots, and surface defects, and the identification results are used as event trigger signals to drive the local compensation module.

[0014] Preferably, the central controller further includes: an online system identification module, a distributed model predictive control module, an adaptive feedforward compensation module, and an anomaly detection and event-driven module; the online system identification module updates the bending stiffness, springback coefficient, and friction parameters of the plate based on real-time sensing data, and the distributed model predictive control module aims to minimize the residual bending energy and stress nonuniformity in the future steps and solves the optimization problem under the constraint of roller force and displacement limits.

[0015] Preferably, the distributed model predictive control module calculates the control quantity using a quadratic programming QP solver or a sparse fast solver in each control cycle, and sends the feedforward displacement, which includes the springback prediction, to the segmented controllable leveling roller unit in advance; the adaptive feedforward compensation module generates compensation instructions for rate and thickness changes based on the prediction results of the laser profile several steps earlier.

[0016] Preferably, the local compensation module includes a rapid impact indenter and a variable stiffness back pressure plate. After visual recognition or laser detection locates a local defect, the anomaly detection and event-driven module triggers the local compensation module to perform pre-compression or local support before the defect arrives, so as to suppress local bulges and reduce disturbance to the overall leveling strategy.

[0017] Preferably, after receiving the plate surface morphology and stress prediction information from the central controller, the traction and shearing unit dynamically adjusts the cutting trigger time and the blade position based on a shearing timing compensation model that includes transmission inertia, blade delay and material elastic rebound. At the same time, force closed-loop control is used at the moment of cutting to reduce blade impact and reduce burrs on the cut.

[0018] Preferably, the central controller maintains a process knowledge base, which records material type, thickness, linear speed, leveling roller settings, local compensation parameters, shearing tool parameters, and corresponding quality index mappings; the central controller updates the online identification model parameters and controller weights based on historical operating data using an incremental learning algorithm to achieve self-learning and process optimization.

[0019] Preferably, the central controller communicates with each sub-controller via a real-time industrial bus, which is either EtherCAT or Profinet, to meet the real-time requirements of online identification and distributed model predictive control.

[0020] The control system also includes a human-machine interface (HMI) to display predicted surface shapes, control commands, defect alarms and process suggestions, and to support manual intervention by operators.

[0021] Preferably, the central controller is equipped with emergency protection logic: when a sensor failure, leveling roller overload, or shearing blade holder overload is detected, the central controller maintains production line stability and issues an alarm by using a predefined safety strategy of reducing line speed, switching to safe traction, or entering bypass mode; the event is also recorded in the process knowledge base for subsequent analysis and model correction.

[0022] Preferably, each section of the segmented controllable leveling roller unit includes at least: a position encoder, a force sensor, a servo drive device, and a servo valve or electro-hydraulic servo actuator capable of rapid response, so as to respond in milliseconds to hundreds of milliseconds according to the optimal displacement and force sequence issued by the central controller and ensure the execution accuracy and response rate required by the distributed model predictive control.

[0023] Compared with the prior art, the beneficial effects of the present invention are:

[0024] 1. This invention uses a segmented model prediction leveling driven by multimodal short-time online identification to construct a short-time dynamic model of physical correlation in real time, predict the surface shape and stress in the next few steps, and finely allocate the control quantities of each roller to actively suppress residual curvature and stress unevenness. This overcomes the problem of failure of empirical or open-loop leveling under changes in material batch, temperature and linear speed, avoids edge deviation and rework caused by downstream morphology rebound due to springback, significantly improves flatness stability, shortens debugging time, reduces reliance on manual experience, and provides more stable input conditions for downstream shearing.

[0025] 2. This invention uses a closed-loop collaborative compensation architecture of leveling and shearing to take surface shape-stress prediction as input for shearing triggering and tool position adjustment, thereby achieving dynamic correction of cutting timing and mechanical compatibility. It eliminates the neglect of elastic rebound and inertial hysteresis under independent shearing control, solves the problems of cutting position drift, blade impact and high burr rate, improves cutting edge positioning accuracy, reduces burrs and tool damage, and improves shearing consistency and yield.

[0026] 3. This invention uses an event-driven local rapid compensation module to apply predetermined local mechanical compensation to the defect location within a very short time scale when a weld, hard spot, or local thickness change is detected. This suppresses local bulges or stress peaks, avoids linewidth fluctuations and debugging complexity caused by large overall adjustments, prevents local defects from causing tool damage or scrap, and improves the system.

[0027] 4. This invention reduces the cost of manual trial and error and downtime during batch and material changes by using a process knowledge base and an online incremental learning closed loop. It also eliminates the high dependence on expert experience during the trial production stage, enabling rapid material changeover and adaptation, continuous performance improvement, reduced manual adjustments, accumulation of enterprise-level industrial knowledge, and improved long-term production line efficiency and consistency. Attached Figure Description

[0028] Figure 1 This is a schematic diagram of the control system framework of the present invention;

[0029] Figure 2 This is a schematic diagram of the operation flow of the control system of the present invention. Detailed Implementation

[0030] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0031] Example 1

[0032] Please see Figure 1and Figure 2 One embodiment of the present invention provides an automatic leveling and shearing control system for medium-thick plate processing, applied to a standard medium-thick plate constant-speed production line:

[0033] Material: Low carbon steel grade; thickness t=5mm, plate width W=1500mm;

[0034] Linear velocity: v = 6 m / min;

[0035] Objective: To achieve a flatness ≤ 0.5mm wave height, a shear position deviation ≤ ±1mm, and a burr rate ≤ 2%;

[0036] Hardware and sensor configuration:

[0037] Laser contour scanner: single unit span width scanning, lateral accuracy ±0.05mm, sampling frequency 1kHz;

[0038] Thickness measurement: Electromagnetic and ultrasonic thickness sensors, accuracy ±0.02mm, update rate 200Hz;

[0039] Distributed strain: 10 strain gauges or single-mode fiber arrays are arranged every 150mm along the plate width, with a sampling rate of 200Hz;

[0040] Leveling rollers: 6 sections (equally spaced), each section equipped with a position encoder and force sensor, response delay (including valve and servo) ≤50ms;

[0041] Local compensator: rapid impact indenter (stroke 3-10mm, response delay ≤30ms);

[0042] Traction roller and shear: multi-axis servo traction, shear tool holder with force sensor (sampling 1kHz), tool position encoder resolution 0.01mm;

[0043] Controller: Real-time industrial PC, EtherCAT bus, MPC solver cycle 50ms.

[0044] Control parameters:

[0045] Sampling period (sensor fusion): T s =10ms is used for rate and force loop; MPC scheduling cycle T MPC =50ms;

[0046] Number of prediction steps: prediction window N p =20 (i.e., 20 × 50 ms = 1.0 s prediction range); control steps N c =5;

[0047] MPC weights: Q k =1.0 (residual curvature weight), Q σ=0.5 (stress non-uniformity weight), R=0.01 (control variation penalty).

[0048] Model identification:

[0049] A simplified linear model (piecewise) for plate bending is identified online using recursive least squares (RLS).

[0050] Model form (linear per segment): y k+1 =φ k θ k +w k ;

[0051] y k+1 The next step is to predict the local curvature vector (discretized segment).

[0052] φ k : Regression vector (including current roll position, roll force, thickness, temperature, speed, etc.);

[0053] θ k : Vector of parameters to be estimated (stiffness correlation coefficient, friction coefficient, etc.);

[0054] W k : Measurement of noise.

[0055] RLS update formula (with forgetting factor λ):

[0056] K k =P k-1 φ k T (λ+φ) k P k-1 φ k T ) -1

[0057] θ k =θ k-1 + (y k -φ k θ k-1 )

[0058] P k =1 / λ(IK k φ k P k-1

[0059] P k : Covariance matrix; Initial P0 = αI (α is large, such as 1×104);

[0060] λ: Forgetting factor, 0.98 in the example;

[0061] K k Kalman gain.

[0062] DMPC objective function (discretized form):

[0063] Solve the following quadratic optimization problem in each MPC cycle:

[0064]

[0065] Limited by:

[0066] u min ≤u k+i ≤u max ;

[0067] Δu min ≤Δu k+i ≤Δu max ;

[0068] U: Future N c The control sequence of each step (displacement and force of each roller section);

[0069] K k+i|k : The predicted curvature vector;

[0070] σ k+i|k : The predicted stress distribution vector;

[0071] Δu k+i =u k+i -u k+i-1 ;

[0072] constraint u min u max Example: Displacement ±10mm, maximum roller force F ma x = 150 kN.

[0073] Timing compensation for shearing:

[0074] Predicted cutoff trigger time correction: t trigger =t nominal -Δt;

[0075] Where Δt=τ comp +Δx rebound / v;

[0076] T nomina Nominal trigger time calculated based on board length;

[0077] T comp System delay compensation, τ comp =τ compute +τ actuator For example, τ compute =50ms, τ actuator =30ms, τ comp=80ms;

[0078] Δx rebound The length of the sheet material that will recover at the moment of cutting, estimated according to the springback model, is calculated from the springback coefficient and prestress.

[0079] v: Linear velocity (mm / s).

[0080] Rebound estimation: Δx rebound =C r ·Δ ·L eff ;

[0081] C r : Empirical rebound coefficient; Δ Local strain release; L eff The effective influence length can be determined by the rigidity of the plate segment and the distance from the cutter position to the nearest leveling roller.

[0082] Operation process:

[0083] Feed inspection and template selection, online RLS identification and update θ k The DMPC runs every 50ms, outputs feedforward displacement and sends it to each roller. In the 10ms step, the force and position loops are updated to ensure closed-loop response. The shearing module adjusts the trigger time based on elastic rebound estimation and enters force closed-loop control at the moment of cutting. Abnormalities are triggered by the event-driven module to provide local compensation or deceleration.

[0084] Example 2

[0085] Please see Figure 1 and Figure 2 One embodiment of the present invention provides an automatic leveling and shearing control system for medium-thick plate processing, prioritizing the linear speed of high-speed thin plates.

[0086] Applicable working conditions:

[0087] Thickness t=3mm, plate width W=2000mm;

[0088] Linear velocity v = 25 m / min;

[0089] Objective: To maintain a shearing positioning error of ≤±2mm at high line speeds, prioritize flatness, and avoid line stoppages.

[0090] Hardware differences (compared to Example 1):

[0091] Increased laser scanning and camera sampling frequency: 500Hz laser full-frame reconstruction rate;

[0092] The leveling roller section is increased to 8 sections to achieve finer distributed force control;

[0093] The MPC cycle is shortened to T.MPC =20ms;

[0094] Servo and drive selection for traction and shearing requires higher bandwidth.

[0095] Control parameters:

[0096] Sensor sampling T s =5ms, MPC cycle 20ms;

[0097] Prediction window N p =30;

[0098] MPC weight adjustment: Increase Q K Prioritize ensuring flatness (Example Q) K =1.5, R=0.02).

[0099] At high speeds, the computational delay to the actuator delay has a significant impact on trigger compensation. Shorter TMPC and faster servo devices ensure timely decision-making.

[0100] Lead time t for defect detection lead =x detect As / v decreases (at the same detection distance), the detection distance for both laser and vision needs to be increased (by moving the laser scanning position forward or adding a front detector).

[0101] For hard spot detection decisions, prioritize bypass shearing or deceleration and perform local indentation to avoid tool damage. Example trigger threshold: sudden strain change Δ >2000με or surface height deviation >1.5mm.

[0102] Example 3

[0103] Please see Figure 1 and Figure 2 The present invention provides an embodiment of an automatic leveling and shearing control system for processing medium-thick plates, specifically designed for low-speed, high-rigidity processing of thick plates.

[0104] Applicable working conditions:

[0105] Thickness t=20mm, plate width W=1800mm;

[0106] Linear velocity v = 1 m / min;

[0107] Objective: To ensure safe shearing, prevent tool breakage, and achieve a smooth finish with uniform stress distribution.

[0108] Hardware differences:

[0109] The leveling roller and drive are designed with high-torque hydraulic servo, with a maximum roller force F. max ≥500kN;

[0110] Local compensators require greater stroke and greater impact force;

[0111] Sensor calibration takes into account the thermoelasticity of thick plates and the inertia of large masses.

[0112] Model and parameters:

[0113] Plate bending stiffness B:

[0114] E: Young's modulus; t: plate thickness; V: Poisson's ratio.

[0115] Residual bending energy:

[0116] K(x): Curvature (1 / m); L: Length of the plate segment under consideration; used to set Q in discretized form in MPC. k Weights.

[0117] Control strategy:

[0118] Due to the large inertia, the constraints on the rate of change of control are increased, limiting... To prevent structural vibration, redundant safety thresholds and overload detection are added, and servo valve pressure and motor current are monitored in real time. Example of an overload threshold: A roller current exceeding 120% of its nominal value for 200ms triggers speed reduction and an alarm.

[0119] Example 4

[0120] Please see Figure 1 and Figure 2 The present invention provides an embodiment of an automatic leveling and shearing control system for medium-thick plate processing, wherein the process for weld seam and hard spot identification and local compensation is as follows:

[0121] Applicable working conditions:

[0122] For any medium-thick plate production line, the focus is on addressing continuous weld seams or hard spots (local protrusions); the line speed example is 6m / min.

[0123] Defect identification:

[0124] Visual camera + CNN: The camera is positioned 3m in front of the feed section, and outputs the defect location x after detecting the weld. def The distance relative to the detection point; the height deviation h of the laser profilometer at that position. def Provide precise numerical values.

[0125] Identification criteria: Weld hard spot determination: Height deviation h def >1.0 mm and width <50 mm; or strain abrupt change Δ >1000με.

[0126] Pre-compensation timing calculation:

[0127] Required lead time to reach the compensation point:

[0128] t lead =x def / v-(τ compute +τ actuator +τ safety )

[0129] x def : Distance from the detection point to the compensation execution point (m); v: Linear velocity (m / s); τ compute : Calculation and communication delay (ms); τ actuator : Actuator action delay (ms); τ safety Safety margin (ms).

[0130] Local compensation command:

[0131] Calculate the preload F pre :F pre =k c ·A contact ·Δh

[0132] K c Local stiffness coefficient (N / mm) 2 This can be obtained through experience or online identification;

[0133] A contact Indenter contact projected area (mm) 2 );

[0134] Δh=h def -h nom Height deviation (mm).

[0135] Indentation timing: at t lead At the start of each step, apply pressure to the local indenter in a closed-loop manner up to F. pre Duration t hold =50ms; If the abnormality persists, switch to traction deceleration and record the event to the process library.

[0136] All defect events are recorded and used as process library data for subsequent incremental learning and updates. c With local compensation strategies.

[0137] Example 5

[0138] Please see Figure 1 and Figure 2 One embodiment of the present invention provides an automatic leveling and shearing control system for processing medium-thick plates, which handles sudden thickness changes:

[0139] Applicable working conditions:

[0140] The sheet material exhibits abrupt changes in thickness, either longitudinally or laterally; the linear speed is moderate, and the goal is to maintain flatness and reduce shear deviation.

[0141] Sensing and Detection:

[0142] The thickness measuring device is placed at the front end of the feed line, detects thickness jump regions, and outputs a thickness profile t(x,y); the thickness information is then incorporated into the RLS regression vector φ. k Update the stiffness parameters.

[0143] Compensation logic:

[0144] If there is a longitudinal mutation (along the length): Before the mutation occurs, modify the stiffness matrix B(x) in the MPC model in advance and add the corresponding feedforward control variable in the prediction window.

[0145] For lateral gradient (along width): differentiated roller position control is given for different lateral positions to minimize the width curvature.

[0146] formula:

[0147] The parameters have been defined in Example 3.

[0148] Specific steps:

[0149] A sudden change in thickness is detected; a new B(x,y) is calculated and the model parameters θ are updated. k MPC uses the updated model to calculate future control quantities, prioritizing local increases or decreases in roller force and displacement along abrupt change sections; the shearing module adjusts the tool feed force and cutting speed in advance and increases the tool position margin to prevent burrs on the cut due to the impact of thickness changes on the cutting force.

[0150] Example 6

[0151] Please see Figure 1 and Figure 2 One embodiment of the present invention provides an automatic leveling and shearing control system for medium-thick plate processing, wherein the rapid influence of new materials and the self-learning of the process library are as follows:

[0152] Applicable working conditions: online adaptation to new steel grades, initial lack of template data, and medium line speed and thickness.

[0153] Self-learning and rapid initialization process:

[0154] Rapid identification phase (cold start): Small window RLS (or least squares) is used to estimate the initial model parameters θ0 from short-term data (e.g., the first 20m of board material); the forgetting factor λ is set to 0.95;

[0155] Conservative control start-up: In the initial identification stage, the MPC weight is set to a conservative configuration (R is increased, and the control amplitude is reduced), and the maximum linear speed and maximum roller force are limited;

[0156] Online optimization phase: As data accumulates, incremental learning is used to update the model and MPC weights, gradually relaxing constraints and achieving standard performance;

[0157] Process library update: When the statistical confidence level reaches the threshold, the template of the material is written into the process library.

[0158] Algorithm details:

[0159] Initial identification segment length: L init =20m; Model parameter confidence threshold: If the parameter covariance P k All diagonal elements are less than 10 -2 If the learning step size is 10, then it is considered convergent; Incremental learning step size: learning rate η=10 -4 .

[0160] When the new material is completely cut for the first time, a quality inspection is automatically performed; if it fails to meet the requirements, it is rolled back to the previous stable template and manual intervention is notified.

[0161] Working principle:

[0162] The system uses multi-modal sensors to perform real-time, high-frequency, full-amplitude sampling of the incoming medium-thick plates in the feeding section. After filtering and fusion with multiple sensors, the raw signals are sent to the online identification module. The module uses recursive least squares or incremental learning methods to estimate the dynamic parameters of the plate in real time, and constructs a dynamic prediction model of the plate surface morphology and stress distribution based on the current perception within a limited prediction window.

[0163] Based on the dynamic model obtained through online identification, the central controller uses piecewise distributed model predictive control as its core, solving an optimization problem within each control cycle: minimizing the residual curvature energy and stress nonuniformity in the next few steps under constraints, while also considering the smoothness of the control input. The calculated optimal displacement and force sequence is sent to each independently driveable leveling roller section via a real-time bus. For local defects identified by vision and laser, the event-driven module triggers the local compensation module to perform instantaneous local mechanical adjustments before the defect reaches the execution point, avoiding fluctuations caused by large-scale overall parameter adjustments.

[0164] The leveling subsystem shares short-term predictions of surface shape and stress with springback estimates in real time with the traction and shearing units. The shearing controller establishes a trigger compensation model based on material elastic springback, transmission inertia and tool holder delay, dynamically corrects the cutting trigger time, tool position and cutting force, and adopts a force closed loop at the moment of cutting to reduce impact and burrs. All operating data is continuously written into the process knowledge base, and incremental learning is used to gradually optimize the identification parameters and MPC weights to achieve rapid adaptation for material changes and long-term process improvement.

[0165] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

Claims

1. An automatic leveling and shearing control system for processing medium-thick plates, characterized in that: The control system includes: A multimodal sensing unit is used to collect real-time data on the board's profile, thickness, temperature, and stress-strain along the length and width of the board. The segmented controllable leveling roller unit consists of several independently driven roller sections with force and position feedback; Local compensation module, used to provide instantaneous mechanical compensation for local protrusions, welds or hard spots; The traction and shearing units are equipped with closed-loop control of position, speed, and force, while the shearing unit has closed-loop control of tool position and tool force. The central controller communicates with each unit via a real-time bus and is configured to: identify the dynamic mechanical model of the board material online based on real-time data from the multimodal sensing unit; calculate and output the optimal displacement and force sequence of each leveling roller unit using a segmented model prediction control within a limited prediction window based on the online identification model; and transmit the predicted information of the board surface morphology and stress distribution to the traction and shearing units for dynamic correction of shearing trigger timing and blade position compensation, thereby realizing closed-loop coordinated control of leveling and shearing.

2. The automatic leveling and shearing control system for medium-thick plate processing according to claim 1, characterized in that: The multimodal sensing unit includes: a laser profile scanner, a thickness measurement device, a distributed optical fiber or strain gauge sensor, an infrared temperature sensor, and a vision camera; the vision camera, in conjunction with a lightweight neural network model, is used to identify and locate welds, hard spots, and surface defects, and the identification results serve as event trigger signals to drive the local compensation module.

3. The automatic leveling and shearing control system for medium-thick plate processing according to claim 1, characterized in that: The central controller further includes: an online system identification module, a distributed model predictive control module, an adaptive feedforward compensation module, and an anomaly detection and event-driven module; the online system identification module updates the bending stiffness, springback coefficient, and friction parameters of the plate based on real-time sensing data; the distributed model predictive control module aims to minimize the residual bending energy and stress nonuniformity in the future steps and solves the optimization problem under the constraint of roller force and displacement limits.

4. The automatic leveling and shearing control system for medium-thick plate processing according to claim 3, characterized in that: The distributed model predictive control module calculates the control quantity using a quadratic programming QP solver or a sparse fast solver in each control cycle, and sends the feedforward displacement, which includes the springback prediction, to the segmented controllable leveling roller unit in advance; the adaptive feedforward compensation module generates compensation instructions for rate and thickness changes based on the prediction results of the laser profile several steps earlier.

5. The automatic leveling and shearing control system for medium-thick plate processing according to claim 2, characterized in that: The local compensation module includes a rapid impact indenter and a variable stiffness back pressure plate. After visual recognition or laser detection locates a local defect, the anomaly detection and event-driven module triggers the local compensation module to perform pre-compression or local support before the defect arrives, so as to suppress local bulges and reduce disturbance to the overall leveling strategy.

6. The automatic leveling and shearing control system for medium-thick plate processing according to claim 1, characterized in that: After receiving plate surface morphology and stress prediction information from the central controller, the traction and shearing unit dynamically adjusts the cutting trigger time and blade position based on a shearing timing compensation model that includes transmission inertia, tool holder delay and material elastic rebound. At the same time, force closed-loop control is used at the moment of cutting to reduce blade impact and reduce burrs on the cut.

7. The automatic leveling and shearing control system for medium-thick plate processing according to claim 1, characterized in that: The central controller maintains a process knowledge base, which records material type, thickness, linear speed, leveling roller settings, local compensation parameters, shearing tool parameters, and corresponding quality index mappings. Based on historical operating data, the central controller uses an incremental learning algorithm to update the online identification model parameters and controller weights to achieve self-learning and process optimization.

8. The automatic leveling and shearing control system for medium-thick plate processing according to claim 1, characterized in that: The central controller communicates with each sub-controller via a real-time industrial bus, which is either EtherCAT or Profinet, to meet the real-time requirements of online identification and distributed model predictive control. The control system also includes a human-machine interface (HMI) to display predicted surface shapes, control commands, defect alarms and process suggestions, and to support manual intervention by operators.

9. An automatic leveling and shearing control system for processing medium-thick plates according to claim 7, characterized in that: The central controller is equipped with emergency protection logic: when a sensor failure, leveling roller overload, or shearing blade holder overload is detected, the central controller maintains production line stability and issues an alarm by using predefined safety strategies such as reducing line speed, switching to safe traction, or entering bypass mode; the event is also recorded in the process knowledge base for subsequent analysis and model correction.

10. An automatic leveling and shearing control system for processing medium-thick plates according to claim 1, characterized in that: Each section of the segmented controllable leveling roller unit includes at least: a position encoder, a force sensor, a servo drive device, and a servo valve or electro-hydraulic servo actuator capable of rapid response, so as to respond in milliseconds to hundreds of milliseconds according to the optimal displacement and force sequence issued by the central controller and ensure the execution accuracy and response rate required for distributed model predictive control.

Images