A method and system for compatible rolling of steel and non-ferrous metal multi-material cold rolling

Abstract

The application discloses a steel and non-ferrous metal multi-material cold rolling compatible rolling method and system, and belongs to the technical field of manufacturing execution systems.The method comprises the following steps: identifying the material quality of a plate in real time through X-ray fluorescence and eddy current conductivity dual-mode sensing; calling pre-stored tension-velocity coupling surface parameters based on the material quality category; collecting tension and timing displacement data by using a high-dynamic tension sensor array and an edge position detector; estimating disturbance through an extended Kalman filter state observer, and generating feedforward compensation and feedback correction by a model predictive controller to dynamically adjust the winding tension instruction.The system solves the tension control lag problem caused by the difference in material physical properties in multi-material cold rolling by introducing an online material quality identification and dynamic tension feedforward compensation cooperative mechanism.The traditional proportional-integral-derivative controller relies on fixed parameters and cannot adapt to the changes in the elastic modulus, yield strength and surface friction coefficient of different materials, resulting in tension fluctuation and causing snake deviation.

Description

Technical Field

[0001] This invention belongs to the field of manufacturing execution system technology, specifically relating to a method and system for cold rolling compatible with multiple materials such as steel and non-ferrous metals. Background Technology

[0002] Cold rolling, a crucial step in sheet metal forming, is widely used in the high-precision processing of steel and non-ferrous metals. Its core objective is to apply pressure to metal strips using rolls at room temperature, achieving thickness reduction, improved surface finish, and optimized mechanical properties. In multi-material co-line production scenarios, different metal materials (such as carbon steel, stainless steel, aluminum, and copper alloys) exhibit significant differences in yield strength, elastic modulus, work hardening characteristics, and surface friction coefficients, resulting in highly nonlinear and heterogeneous dynamic responses to rolling tension. Tension control, as a core means of maintaining stable strip operation and preventing deviation and wrinkling, directly determines the quality of the finished sheet shape and the continuity of the production line.

[0003] Tension control in the coiling section is particularly critical, requiring real-time matching with changes in the material's instantaneous deformation resistance and elongation during high-speed operation. Traditional cold rolling systems generally employ PID controllers based on fixed parameters for closed-loop tension regulation, with control logic relying on preset material types and static mechanical models. However, when the production line switches to different materials or processes alloy strips with fluctuating compositions, the fixed parameters cannot adapt to the abrupt changes in the material's dynamic characteristics, leading to lag or overshoot in the tension response. This lag effect accumulates rapidly during high-speed rolling, causing the strip to slither laterally at the coiler inlet, which in severe cases results in edge scratches, coil collapse, or even strip breakage and shutdown.

[0004] While feedforward compensation or segmented parameter tables are introduced to handle multi-material switching, they still rely on manual experience to set thresholds and switching logic, lacking the ability to perceive the real-time strain state of materials. Especially in situations without clear material identification or with slight variations in batch composition, the system struggles to autonomously determine material characteristics, leading to delays in parameter reconstruction. Furthermore, frequent material switching forces operators to repeatedly stop the machine to adjust PID parameters, significantly reducing overall equipment efficiency (OEE) and increasing the proportion of unplanned downtime. Summary of the Invention

[0005] This invention provides a method and system for multi-material cold rolling of steel and non-ferrous metals, aiming to solve the problem of serpentine deviation of sheet metal in the coiling section caused by tension control lag during multi-material cold rolling, and to overcome the technical defects of traditional proportional-integral-derivative controllers, such as frequent downtime for parameter adjustment and low production efficiency due to parameter fixation. This invention constructs an adaptive tension collaborative control architecture based on real-time material identification and dynamic tension feedforward compensation, achieving precise closed-loop tension control for different material sheets during high-speed continuous cold rolling, thereby eliminating serpentine deviation and improving production line compatibility and operational stability.

[0006] This invention provides a method for cold rolling compatible steel and non-ferrous metals, comprising: An online material identification unit is installed at the inlet side of the cold rolling mill to obtain the material category information of the sheet to be rolled in real time. Based on the material category information, the initial tension setting value, rolling speed limit value and coiling tension gradient coefficient corresponding to the current plate are retrieved from the preset material-process parameter mapping database. During the rolling process, a high dynamic response tension sensor array installed between the uncoiler, intermediate roll system and coiler is used to collect tension time sequence data of each key section in real time; the lateral displacement deviation signal output by the plate edge position detector is acquired simultaneously. The tension time series data and the lateral displacement deviation signal are input into a multivariable state observer to generate a tension error state vector characterizing the dynamic disturbance of the system. Based on the tension error state vector, the current material type, and the real-time rolling speed, the feedforward compensation and feedback correction are calculated by the online updated model prediction controller; the feedforward compensation is then superimposed on the initial tension setpoint to form a dynamic tension command. The dynamic tension command is sent to the vector frequency converter driver of the winding main motor to adjust the winding torque and achieve millisecond-level closed-loop control of tension.

[0007] In one embodiment of the present invention, the online material identification unit includes an X-ray fluorescence spectrometer and an eddy current conductivity detection probe, which are symmetrically arranged above the inlet guide roller along the width direction of the plate, with a sampling frequency of not less than 20 times per second; the X-ray fluorescence spectrometer is used to determine the elemental composition ratio of the plate surface, and the eddy current conductivity detection probe is used to measure the electromagnetic response characteristics of the plate; the material category information is output by a Bayesian classifier that fuses the above two types of sensor data, and the classification results include four standard material identifiers: low carbon steel, high strength steel, aluminum alloy, and copper alloy.

[0008] As one embodiment of the present invention, the material-process parameter mapping database pre-stores a tension-speed coupling relationship surface corresponding to each type of standard material identifier. This surface is fitted with a cubic B-spline surface with rolling speed as the abscissa, target tension as the ordinate, and coiling acceleration as the third dimension. The initial tension setting value is obtained by interpolation of the current real-time rolling speed on the surface. The coiling tension gradient coefficient is defined as the tension adjustment slope caused by a unit speed change, and its value ranges from 0.5 to 3.5 kN of tension adjustment per meter per second of speed change.

[0009] As one embodiment of the present invention, the high dynamic response tension sensor array includes three independently installed magnetic tension detection rollers, located at the exit of the first frame, the entrance of the last frame, and the entrance of the winding machine, respectively; each detection roller integrates eight orthogonally arranged magnetostrictive strain sensitive elements, with a sampling period of 10 milliseconds; the tension time series data is an effective tension value sequence after temperature drift compensation and nonlinear correction.

[0010] In one embodiment of the present invention, the plate edge position detector adopts a dual-sided laser triangulation distance sensor, which is installed on a fixed bracket two meters in front of the winding machine, with a measurement accuracy of ±0.1 mm; the lateral displacement deviation signal is the steady-state component of the difference between the distance values ​​on the left and right sides after low-pass filtering.

[0011] As one embodiment of the present invention, the multivariable state observer adopts an extended Kalman filter structure, and its state equation is obtained by discretization of the electromechanical coupling dynamics model of the cold rolling mill. The state variables include the actual tension of the coiling section, the rate of change of tension, and unknown external disturbance terms. The observer gain matrix is ​​updated online according to the inertia parameter and damping coefficient corresponding to the current material type, and the update period is synchronized with the sampling period of the tension sensor.

[0012] In one embodiment of the present invention, the model prediction controller operates in a rolling time-domain optimization mode, with a prediction time domain length of 500 milliseconds and a control time domain length of 200 milliseconds; its cost function includes three weighted terms: the squared term of tension tracking error, the squared term of control increment change rate, and the squared term of lateral displacement deviation; the weight coefficients are automatically configured according to the material type, wherein the lateral displacement deviation weight corresponding to aluminum alloy material is 2.5 times that of low carbon steel material.

[0013] In one embodiment of the present invention, the feedforward compensation amount is calculated by the model predictive controller based on the tension error trend in the predicted time domain, and its expression is the convolution result of the tension error state vector at the current moment and the preset disturbance suppression transfer function; the feedback correction amount is the first term of the optimal control increment sequence output by the model predictive controller; the dynamic tension command is equal to the initial tension setpoint plus the feedforward compensation amount plus the feedback correction amount.

[0014] As one embodiment of the present invention, during the rolling switching stage, when the online material identification unit detects a change in the material category of the sheet, the system automatically triggers a parameter smoothing transition mechanism. This mechanism switches the tension-velocity surface parameters corresponding to the old material to the parameters corresponding to the new material through linear interpolation within 0.5 seconds, while freezing the weight coefficient update of the model prediction controller until the new material is stably rolled for more than 3 seconds and then resumes adaptive adjustment.

[0015] This invention provides a multi-material cold rolling compatible rolling system for steel and non-ferrous metals, comprising: The online material identification unit is located at the inlet side of the cold rolling mill and is used to output the material category information of the sheet to be rolled in real time. The material-process parameter mapping database stores tension-velocity coupling relationship surfaces and winding tension gradient coefficients corresponding to various standard material identifiers; A high dynamic response tension sensor array is distributed in the key section between the uncoiler and the coiler to collect tension timing data; a sheet edge position detector is installed in front of the coiler to output lateral displacement deviation signals. A multivariable state observer receives the tension timing data and the lateral displacement deviation signal, and outputs a tension error state vector. The model prediction controller receives the tension error state vector, material category information and real-time rolling speed, and outputs the feedforward compensation amount and the feedback correction amount. The dynamic tension command generation module superimposes the initial tension setpoint, feedforward compensation, and feedback correction to form a dynamic tension command. The main winding motor vector frequency converter driver receives the dynamic tension command and adjusts the winding torque to achieve closed-loop tension control.

[0016] In one embodiment of the present invention, the online material identification unit includes an X-ray fluorescence spectrometer and an eddy current conductivity detection probe. The data fusion of the two is completed by a Bayesian classifier, and four types of standard material identifiers are output.

[0017] As one embodiment of the present invention, the high dynamic response tension sensor array consists of three piezomagnetic tension detection rollers, each of which has eight magnetostrictive strain sensitive elements built in, and has temperature drift compensation and nonlinear correction functions.

[0018] In one embodiment of the present invention, the multivariable state observer adopts an extended Kalman filter structure, and its state equation is based on the electromechanical coupling dynamics model of the cold rolling mill. The observer gain matrix is ​​updated online according to the inertia and damping parameters corresponding to the current material.

[0019] As one embodiment of the present invention, the model predictive controller operates in a rolling time-domain optimization mode, and its cost function includes three weighted terms: tension tracking error, control increment change rate, and lateral displacement deviation. The weighting coefficients are automatically configured according to the material type.

[0020] As one embodiment of the present invention, the system further includes a parameter smoothing transition module, which performs linear interpolation switching within 0.5 seconds when switching materials, and freezes the model prediction controller weight update until the new material is stably rolled for more than 3 seconds.

[0021] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. This invention fundamentally solves the problem of tension control lag caused by differences in the physical properties of materials in multi-material cold rolling by introducing a collaborative mechanism of online material identification and dynamic tension feedforward compensation. Traditional proportional-integral-derivative controllers rely on fixed parameters and cannot adapt to changes in the elastic modulus, yield strength, and surface friction coefficient of different materials, leading to tension fluctuations and serpentine deviations.

[0022] 2. This invention utilizes dual-modal sensing of X-ray fluorescence and eddy current conductivity to achieve millisecond-level material discrimination. Combined with a pre-stored tension-velocity coupling surface, it ensures a high degree of matching between the initial tension setpoint and the current material. Simultaneously, a multi-source feedback mechanism is constructed using a high-dynamic tension sensor array and an edge position detector to drive an extended Kalman state observer to estimate system disturbances in real time. A model predictive controller then generates dynamic tension commands with both feedforward suppression and feedback correction capabilities, reducing the tension control response time to less than 50 milliseconds.

[0023] 3. This invention completely eliminates the serpentine deviation phenomenon in the coiling section, enabling continuous rolling of multiple materials without stopping for parameter adjustment, improving production line switching efficiency by more than 40%, and controlling the steady-state error of tension control within ±2% of the target value, significantly improving the quality of finished product shape and equipment utilization. Attached Figure Description

[0024] Figure 1 This is a schematic diagram of the overall technical solution architecture of a multi-material cold rolling compatible rolling method and system for steel and non-ferrous metals proposed in this invention. Figure 2 This is a schematic diagram of the core principle framework of the adaptive tension collaborative control architecture based on the combination of real-time material recognition and dynamic tension feedforward compensation in this invention. Figure 3 This is a flowchart illustrating the logical flow of the online material identification and initial process parameter retrieval stage in this invention. Figure 4 This is a flowchart illustrating the logical process of the tension time-series data and lateral displacement deviation fusion processing and tension error state vector generation stages in this invention. Figure 5 This is a flowchart illustrating the logical flow of the model predictive controller in this invention, which generates feedforward compensation and feedback correction quantities to form dynamic tension commands. Figure 6 This is a schematic diagram of the interaction relationship and data flow of the multi-level sensing, control and drive units between the cold rolling mill inlet and the coiler in this invention. Detailed Implementation

[0025] Please refer to Figures 1 to 6 This invention provides a method and system for multi-material cold rolling compatible with steel and non-ferrous metals. It aims to solve the problem of serpentine deviation in the coiling section caused by tension control lag during multi-material cold rolling, and overcomes the technical defects of traditional proportional-integral-derivative controllers, such as frequent downtime for parameter adjustment and low production efficiency due to parameter fixation. This method constructs an adaptive tension collaborative control architecture based on real-time material identification and dynamic tension feedforward compensation, achieving precise closed-loop tension control for different material plates during high-speed continuous cold rolling, thereby eliminating serpentine deviation and improving production line compatibility and operational stability.

[0026] The method includes the following steps: S1, An online material identification unit is set at the inlet side of the cold rolling mill to obtain the material category information of the sheet to be rolled in real time; S2, based on the material category information, retrieve the initial tension setting value, rolling speed upper limit value and coiling tension gradient coefficient corresponding to the current plate from the preset material-process parameter mapping database; S3, during the rolling process, uses a high dynamic response tension sensor array installed between the uncoiler, intermediate roll system and coiler to collect tension timing data of each key section in real time; S4, synchronously acquire the lateral displacement deviation signal output by the plate edge position detector; S5, input the tension time series data and the lateral displacement deviation signal to the multivariable state observer to generate a tension error state vector characterizing the dynamic disturbance of the system; S6, based on the tension error state vector, the current material type and the real-time rolling speed, the feedforward compensation amount and the feedback correction amount are calculated by the online updated model prediction controller; S7, the feedforward compensation is superimposed on the initial tension setting value to form a dynamic tension command; S8, the dynamic tension command is sent to the vector frequency converter driver of the winding main motor to adjust the winding torque and realize millisecond-level closed-loop control of tension.

[0027] In step S1, the online material identification unit includes an X-ray fluorescence spectrometer and an eddy current conductivity detector, both symmetrically arranged above the inlet guide roller along the width of the plate, with a sampling frequency of no less than 20 times per second. The X-ray fluorescence spectrometer is used to determine the elemental composition ratio of the plate surface. Its excitation source is a rhodium target X-ray tube, with an operating voltage of 45 kV and a current of 1.5 mA. The detector is a silicon drift detector with an energy resolution of 130 eV, capable of simultaneously detecting the characteristic X-ray intensities of six major elements: iron, aluminum, copper, manganese, chromium, and nickel. The eddy current conductivity detector employs a differential dual-coil structure, with an excitation frequency of 500 kHz, a coil diameter of 25 mm, and a lift-off distance controlled within 0.5 mm. By measuring the amplitude and phase angle changes of the induced electromotive force, the electromagnetic response characteristics of the plate are inverted. After the two types of sensor data are timestamped and aligned, they are input into a Bayesian classifier for fusion and discrimination.

[0028] The prior probabilities of the Bayesian classifier are derived from historical rolling batch statistics. The likelihood function is constructed based on a Gaussian distribution model of four standard material identifiers (low-carbon steel, high-strength steel, aluminum alloy, and copper alloy) in a two-dimensional feature space of element ratio and electrical conductivity. The classification results are output in discrete integer encoding form, with the encoding rule: one represents low-carbon steel, two represents high-strength steel, three represents aluminum alloy, and four represents copper alloy. This identification process completes a full judgment every 10 milliseconds, ensuring that the material information update rate matches the subsequent control cycle.

[0029] In step S2, the material-process parameter mapping database pre-stores the tension-velocity coupling surface corresponding to each type of standard material identifier. This surface is fitted using a cubic B-spline surface with rolling speed as the abscissa, target tension as the ordinate, and coiling acceleration as the third dimension. The cubic B-spline basis functions are defined in the velocity range of 0 to 30 m / s and the acceleration range of 0 to... In the parameter domain, the number of control vertices is 7×5, and each control vertex corresponds to a measured tension value, which is obtained through historical rolling tests under steady-state conditions. The initial tension setpoint is obtained by bilinear interpolation of the current real-time rolling speed and the current coiling acceleration on the surface.

[0030] The coiling tension gradient coefficient is defined as the tension adjustment slope caused by a unit speed change, with a value ranging from 0.5 to 3.5 kN of tension adjustment per meter per second of speed change. For low-carbon steel, this coefficient is 1.2 kN / m / s; for high-strength steel, it is 2.8 kN / m / s; for aluminum alloys, it is 0.7 kN / m / s; and for copper alloys, it is 1.5 kN / m / s. Furthermore, the database stores the upper limit of rolling speed for each material type: 25 meters per second for low-carbon steel, 18 meters per second for high-strength steel, 22 meters per second for aluminum alloys, and 20 meters per second for copper alloys. When the real-time rolling speed exceeds this upper limit, the system automatically triggers a speed reduction protection logic to prevent tension loss of control due to overspeed.

[0031] In step S3, the high dynamic response tension sensor array includes three independently installed piezomagnetic tension detection rollers, located at the exit of the first frame, the entrance of the last frame, and the entrance of the winding machine, respectively. Each detection roller has a diameter of 300 mm and is covered with a 5 mm thick polyurethane elastic layer to reduce scratches on the sheet surface. Inside each detection roller are eight orthogonally arranged magnetostrictive strain-sensitive elements, distributed in a ring at 45-degree intervals. Each element consists of a permalloy sheet and a miniature Hall sensor, with a sensitivity of 0.5 mV per micro-strain. The raw signals from the eight channels are pre-amplified, low-pass filtered, and then sent to a 16-bit analog-to-digital converter with a sampling period of 10 milliseconds. Tension calculation uses a weighted average algorithm, with weights dynamically adjusted according to the signal-to-noise ratio of each channel. The raw tension value needs to undergo temperature drift compensation and nonlinear correction. Temperature drift compensation is based on a built-in platinum resistance thermometer, and the compensation formula is: ; in, To compensate for the post-tension value, This is the original tension value. For ambient temperature, The reference temperature for calibration is 25 degrees Celsius. The temperature drift coefficient is retrieved from the database based on the material type. Nonlinear correction employs piecewise cubic spline interpolation, with 11 correction nodes evenly distributed between 0 and 50 kN. The final output tension time-series data is a sequence of effective tension values ​​after double correction, with the time resolution consistent with the sampling period.

[0032] In step S4, the plate edge position detector uses a dual-sided laser triangulation distance sensor, mounted on a fixed bracket two meters in front of the winding machine. The left and right sensors are symmetrically arranged on either side of the plate's centerline, with a measurement spot diameter of 0.3 mm and a sampling frequency of 1000 Hz. The raw distance values ​​output by the sensors are filtered by a moving average (window length of 50 points), and the difference between the left and right distance values ​​is calculated to obtain the lateral offset of the plate's centerline relative to the reference position. This offset is then processed by a second-order Butterworth low-pass filter with a cutoff frequency of 5 Hz to filter out high-frequency vibration interference, retaining the steady-state component as the lateral displacement deviation signal. The dynamic range of this signal is ±10 mm, and the resolution is better than 0.05 mm, meeting the accuracy requirements for serpentine offset detection.

[0033] In step S5, the multivariable state observer employs an extended Kalman filter structure. Its state equation is obtained by discretizing the electromechanical coupling dynamics model of the cold rolling mill, with a discretization step size equal to the sampling period of the tension sensor (10 milliseconds). The state variables include the actual tension of the winding section. rate of change of tension and unknown external disturbance terms This constitutes a three-dimensional state vector. State transition matrix Derived from the following system of differential equations: ; in, The moment of inertia of the drum. The viscous damping coefficient is... Where is the radius of the drum. For the motor output torque, Let be the angular velocity. After discretization, the state transition function is of non-linear form, containing the current material's corresponding angular velocity. and Parameters. The observation equation is: ; Among them, the observation vector , This is the measurement value from the tension sensor at the inlet of the winding machine. This is a lateral displacement deviation signal. For the observation matrix, For observation noise. Observer gain matrix. The inertia parameters and damping coefficients corresponding to the current material type are updated online, with the update cycle synchronized with the tension sensor sampling cycle. In each iteration, a time update is performed first, followed by a measurement update, ultimately outputting the tension error state vector. ,in This constitutes the principal component of the tension tracking error.

[0034] In step S6, the model prediction controller operates in a rolling time-domain optimization mode, with a prediction time domain length of 500 milliseconds, corresponding to 50 sampling steps; and a control time domain length of 200 milliseconds, corresponding to 20 sampling steps. Its cost function... Defined as: ; in, For the first Predicting tension step by step The reference tension (i.e., the initial tension setting value). For the first Predicting lateral displacement deviation step by step. To control the increment (i.e. the change in tension command). , , This refers to the weighting coefficient. The weighting coefficient is automatically configured based on the material type: for low-carbon steel, =1.0, =0.4, =0.1; For high-strength steel, =1.0, =0.6, =0.15; for aluminum alloys, =1.0, =1.0, =0.08; for copper alloys, =1.0, =0.7, =0.12. The optimization problem is solved within each control cycle to obtain the optimal control increment sequence. Feedforward compensation The expression is calculated from the convolution result of the tension error state vector and the preset perturbation suppression transfer function: ; in, It is an exponentially decaying impulse response function with a time constant. Set according to material type. Feedback correction amount. This is the first term of the optimal control increment sequence.

[0035] In step S7, the dynamic tension command equal to the initial tension setting value Add feedforward compensation Plus feedback corrections ,Right now: ; The command is processed by a rate limiter to prevent sudden changes in command from impacting the mechanical system. The rate limit threshold is set according to the current material: 500 N / s for low-carbon steel, 300 N / s for high-strength steel, 400 N / s for aluminum alloy, and 350 N / s for copper alloy.

[0036] In step S8, the dynamic tension command is converted into a target torque command for the winding main motor. The conversion relationship is based on real-time roll diameter calculation, which is derived from the linear velocities of the uncoiler and winding machine and the cumulative rotation angle integral. The target torque command is sent to the vector frequency converter driver via the CAN bus. The driver adopts a field-oriented control strategy with a current loop bandwidth of not less than 1000 Hz to ensure a torque response delay of less than 5 milliseconds. The actual torque output by the driver forms an inner closed loop through encoder feedback, and the outer loop tension control achieves millisecond-level dynamic response by adjusting the torque command.

[0037] During the rolling changeover phase, when the online material identification unit detects a change in the sheet material type, the system automatically triggers a parameter smoothing transition mechanism. This mechanism linearly interpolates the tension-velocity surface parameters corresponding to the old material to those corresponding to the new material within 0.5 seconds, with an interpolation step size of 10 milliseconds and a total of 50 steps. During this period, the weight coefficients of the model predictive controller are frozen to the old material configuration to prevent control oscillations due to sudden parameter changes. Simultaneously, feedforward compensation calculations are paused, relying solely on feedback corrections to maintain basic stability. Once the new material has been stably rolled for more than 3 seconds, the system resumes adaptive adjustments, including weight coefficient updates, feedforward compensation restart, and observer gain reset.

[0038] The system includes an online material identification unit, a material-process parameter mapping database, a high dynamic response tension sensor array, a sheet metal edge position detector, a multivariable state observer, a model predictive controller, a dynamic tension command generation module, and a vector frequency converter driver for the winding main motor. The online material identification unit outputs material category information to the database and the model predictive controller; the database outputs the initial tension setpoint to the dynamic tension command generation module; the tension sensor array and edge position detector output sensor data to the multivariable state observer; the state observer outputs the tension error state vector to the model predictive controller; the model predictive controller outputs feedforward compensation and feedback correction to the dynamic tension command generation module; the dynamic tension command generation module outputs the dynamic tension command to the vector frequency converter driver; and the vector frequency converter driver drives the winding main motor to perform tension control. All modules are interconnected via industrial real-time Ethernet with a communication cycle of 10 milliseconds to ensure data synchronization and timing consistency.

[0039] The system also includes a parameter smoothing transition module, which takes over the database output and controller configuration when a material switching event is triggered, executes linear interpolation and freeze logic, and releases control after a 3-second countdown. During system operation, all key variables are recorded in a historical database for subsequent process optimization and fault diagnosis.

[0040] Using the above methods and systems, the tension control response time in the multi-material cold rolling process is shortened to less than 50 milliseconds, the tension steady-state error is controlled within ±2% of the target value, the snake-like offset phenomenon in the coiling section is completely eliminated, and continuous multi-material rolling without stopping the machine to adjust parameters is achieved, improving the production line switching efficiency by more than 40%.

Claims

1. A method for cold rolling compatible with multiple materials including steel and non-ferrous metals, characterized in that, include: An online material identification unit is installed at the inlet side of the cold rolling mill to obtain the material category information of the sheet to be rolled in real time. Based on the material category information, the initial tension setting value, rolling speed limit value and coiling tension gradient coefficient corresponding to the current plate are retrieved from the preset material-process parameter mapping database. During the rolling process, a high dynamic response tension sensor array installed between the uncoiler, intermediate roll system and coiler is used to collect tension time sequence data of each key section in real time. Synchronously acquire the lateral displacement deviation signal output by the plate edge position detector; The tension time series data and the lateral displacement deviation signal are input into a multivariable state observer to generate a tension error state vector characterizing the dynamic disturbance of the system. Based on the tension error state vector, the current material type, and the real-time rolling speed, the feedforward compensation and feedback correction are calculated by the online updated model prediction controller. The feedforward compensation is superimposed on the initial tension setting value to form a dynamic tension command; The dynamic tension command is sent to the vector frequency converter driver of the winding main motor to adjust the winding torque and achieve millisecond-level closed-loop control of tension.

2. The multi-material cold rolling compatible rolling method for steel and non-ferrous metals according to claim 1, characterized in that, An online material identification unit is installed at the inlet side of the cold rolling mill to obtain real-time material category information of the sheet metal to be rolled, including: The X-ray fluorescence spectrometer and the eddy current conductivity detection probe are symmetrically arranged above the inlet guide roller along the width of the plate, and the sampling frequency is not less than 20 times per second. The X-ray fluorescence spectrometer was used to determine the elemental composition ratio of the plate surface, and the eddy current conductivity detection probe was used to measure the electromagnetic response characteristics of the plate. The two types of sensor data are aligned with timestamps and then input into a Bayesian classifier for fusion and discrimination, outputting four standard material identifiers as the material category information.

3. The multi-material cold rolling compatible rolling method for steel and non-ferrous metals according to claim 2, characterized in that, Based on the material category information, the initial tension setting value, rolling speed upper limit value, and coiling tension gradient coefficient corresponding to the current sheet are retrieved from the preset material-process parameter mapping database, including: The tension-velocity coupling surface corresponding to the current material category is read from the material-process parameter mapping database. This surface is fitted with a cubic B-spline surface with rolling speed as the horizontal axis, target tension as the vertical axis, and coiling acceleration as the third dimension. The initial tension setting value is obtained by bilinear interpolation on the curved surface based on the current real-time rolling speed and coiling acceleration; Read the winding tension gradient coefficient corresponding to the current material type. Its value ranges from 0.5 to 3.5 kN of tension adjustment per meter per second of speed change.

4. The multi-material cold rolling compatible rolling method for steel and non-ferrous metals according to claim 3, characterized in that, During the rolling process, a high dynamic response tension sensor array installed between the uncoiler, intermediate roll system, and coiler collects real-time tension timing data for each key section, including: Three magnetic tension detection rollers are installed at the exit of the first frame, the entrance of the last frame, and the entrance of the winding machine, respectively. Each magnetic tension sensing roller integrates eight orthogonally arranged magnetostrictive strain sensing elements with a sampling period of 10 milliseconds; After temperature drift compensation and nonlinear correction are performed on the original tension signal, the effective tension value sequence is output as the tension time series data.

5. The multi-material cold rolling compatible rolling method for steel and non-ferrous metals according to claim 4, characterized in that, Synchronously acquire the lateral displacement deviation signal output by the plate edge position detector, including: The dual-sided laser triangulation rangefinder was installed on a fixed bracket two meters in front of the winding machine, with a measurement accuracy of ±0.1 mm; The difference between the distance measurements on the left and right sides is calculated, and the steady-state component is extracted after low-pass filtering as the lateral displacement deviation signal.

6. The multi-material cold rolling compatible rolling method for steel and non-ferrous metals according to claim 5, characterized in that, The tension time-series data and lateral displacement deviation signal are input to a multivariable state observer to generate a tension error state vector characterizing the dynamic disturbance of the system, including: Construct a three-dimensional state vector with the actual tension of the winding segment, the rate of change of tension, and unknown external disturbance terms as state variables; The nonlinear state transition function is obtained by discretizing the electromechanical coupling dynamics model of the cold rolling mill. The observer gain matrix is ​​updated online based on the inertia parameters and damping coefficients corresponding to the current material type. The time update and measurement update steps of extended Kalman filtering are performed to output the tension error state vector.

7. The multi-material cold rolling compatible rolling method for steel and non-ferrous metals according to claim 6, characterized in that, Based on the tension error state vector, the current material type, and the real-time rolling speed, the feedforward compensation and feedback correction are calculated by the online-updated model prediction controller, including: The model predictive controller is run in a rolling time-domain optimization mode, with a prediction time domain length of 500 milliseconds and a control time domain length of 200 milliseconds. A cost function is constructed that includes the squared term of tension tracking error, the squared term of control increment change rate, and the squared term of lateral displacement deviation, with the weighting coefficients automatically configured according to the material type; Solve for the optimal control increment sequence and take its first term as the feedback correction amount; The feedforward compensation amount is obtained by convolving the tension error state vector with a preset disturbance suppression transfer function.

8. The method for cold rolling compatible steel and non-ferrous metals of multiple materials according to claim 7, characterized in that, The feedforward compensation is superimposed on the initial tension setpoint to form a dynamic tension command, including: The initial tension setting value, the feedforward compensation amount, and the feedback correction amount are added together to obtain the unlimited dynamic tension command; A material-related rate limiting process is applied to the unlimited dynamic tension command to generate the final dynamic tension command.

9. The multi-material cold rolling compatible rolling method for steel and non-ferrous metals according to claim 8, characterized in that, During the rolling changeover phase, when the online material identification unit detects a change in the material type of the sheet metal, the system automatically triggers a parameter smoothing transition mechanism, including: Within 0.5 seconds, the linear interpolation of the tension-velocity coupling surface parameters corresponding to the old material is switched to the parameters corresponding to the new material; Freeze the update of the weight coefficients of the model predictive controller and pause the calculation of the feedforward compensation. After the new material has been stably rolled for more than 3 seconds, the adaptive adjustment function is restored.

10. A multi-material cold rolling compatible rolling system for steel and non-ferrous metals, characterized in that, include: The online material identification unit is located at the inlet side of the cold rolling mill and is used to output the material category information of the sheet to be rolled in real time. The material-process parameter mapping database stores tension-velocity coupling relationship surfaces and winding tension gradient coefficients corresponding to various standard material identifiers; A high dynamic response tension sensor array is distributed in the key section between the uncoiler and the winding machine to collect tension time-series data; The sheet edge position detector is installed in front of the coiler and is used to output lateral displacement deviation signals; A multivariable state observer receives the tension timing data and the lateral displacement deviation signal, and outputs a tension error state vector. The model prediction controller receives the tension error state vector, material category information and real-time rolling speed, and outputs the feedforward compensation amount and the feedback correction amount. The dynamic tension command generation module superimposes the initial tension setpoint, feedforward compensation, and feedback correction to form a dynamic tension command. The main winding motor vector frequency converter driver receives the dynamic tension command and adjusts the winding torque to achieve closed-loop tension control.

Images