Method for controlling the cutting of steel strips for advanced ultra-high strength sheets and plated sheets thereof

Abstract

The application discloses a kind of advanced ultra-high strength plate and its steel strip transverse cutting control method for plate processing with coating layer.The method is based on geometric parameters, material characteristic parameters, tool parameters and process adjustment parameters to construct steel strip transverse cutting digital twin model;In the shearing production process, shear force signal, vibration signal and temperature field distribution data are collected, and a finite element simulation model is established based on the digital twin model and real-time working condition data;Extracting shear quality characteristic parameters;Using deep neural network algorithm to establish quality prediction model between process parameters and shear quality characteristic parameters;Based on the quality prediction model, a plurality of parameter combinations are simulated and calculated in the process adjustment parameter range, and the optimal process parameter combination is selected as the recommended parameter;According to the recommended parameter, transverse cutting equipment is configured.The application breaks through the dependence of traditional process parameter setting on operator experience, realizes the theoretical and quantitative control of shearing quality, and improves product quality consistency and production efficiency.

Description

Technical Field

[0001] This invention belongs to the field of metal material processing, specifically relating to a method for controlling the cross-cutting of steel strips used in the processing of advanced ultra-high strength plates and their coated plates. Background Technology

[0002] As a vital foundation of the national economy, the steel industry's products are widely used in automobile manufacturing, construction machinery, shipbuilding, and marine engineering. With downstream industries continuously raising their requirements for material performance, advanced ultra-high-strength steel plates and their coated plates, due to their superior mechanical properties, corrosion resistance, and lightweight advantages, have become core raw materials for high-end equipment manufacturing. Steel strip cross-cutting, a crucial process in steel deep processing, directly impacts the quality stability, dimensional consistency, and production efficiency of the final product through its processing precision and control level. Therefore, the research and optimization of steel strip cross-cutting control technology has always been an important topic in the steel processing field.

[0003] The steel strip cross-cutting control method involves multiple process steps, including setting shearing force parameters, adjusting tool gap, controlling tension, and optimizing shearing speed. Its core objective is to achieve high-precision, high-efficiency, and high-stability shearing and forming. Traditional cross-cutting control mainly relies on manual experience to set process parameters and uses offline detection to evaluate the shearing quality afterward. This extensive control mode is difficult to adapt to the characteristics of advanced ultra-high strength steel plate materials, such as large performance fluctuations, narrow processing windows, and high sensitivity to process parameters. Especially when dealing with steel plates of different strength levels and thicknesses, it is often necessary to frequently change tooling dies and repeatedly adjust parameters, which not only leads to low production efficiency but also easily causes quality problems such as shearing burrs, dimensional deviations, and poor shapes.

[0004] Existing technologies for controlling the cross-cutting of steel strips have several shortcomings: First, the setting of process parameters lacks theoretical guidance and data support, relying excessively on the subjective experience of operators, resulting in poor product quality consistency and large batch-to-batch fluctuations. Second, the shearing process lacks a real-time monitoring and dynamic adjustment mechanism, making it impossible to respond immediately to interference factors such as changes in material properties and tool wear. Third, there is a lack of forward-looking prediction capabilities for shearing results, making it impossible to predict the shearing quality under specific process parameters before production and optimize accordingly. These problems are particularly prominent in the processing of advanced ultra-high strength steel plates and their coated plates, urgently requiring a steel strip cross-cutting control method that can simulate the shearing process and predict optimized parameters to meet the production demands for high precision, high efficiency, and high stability. Summary of the Invention

[0005] The purpose of this invention is to provide a method for controlling the cross-cutting of steel strips used in the processing of advanced ultra-high strength plates and their coated plates, which can effectively solve the problems in the background art.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0007] This invention provides a method for controlling the cross-cutting of steel strips used in the processing of advanced ultra-high strength steel plates and their coated plates, comprising the following steps:

[0008] S1. Based on the obtained geometric parameters, material property parameters, tool parameters, and process adjustment parameters of the steel strip, a digital twin model of steel strip cross-cutting is constructed as process parameters;

[0009] S2. Collect shear force signals, vibration signals, and temperature field distribution data during the shearing production process to construct a real-time working condition dataset;

[0010] S3. Based on the digital twin model and real-time working condition data, a finite element simulation model is established, and the stress field distribution, strain field distribution and temperature field distribution of the shear region are obtained through simulation calculation;

[0011] S4. Extract the shear burr height, shear collapse angle depth, cross-sectional roughness, and dimensional accuracy as shear quality characteristic parameters;

[0012] S5. Based on historical production data, a quality prediction model between process parameters and shear quality characteristic parameters is established using machine learning algorithms;

[0013] S6. Based on the quality prediction model, perform simulation calculations on the range of process adjustment parameters using multiple parameter combinations, and select the optimal process parameter combination as the recommended parameters;

[0014] S7. Configure the cross-cutting equipment according to the recommended parameters, continuously collect real-time working condition data during the shearing production process and input it into the quality prediction model for real-time prediction. When the prediction result deviates from the preset quality threshold, the process parameter adjustment command is automatically triggered to realize dynamic optimization control of the shearing process.

[0015] According to one embodiment of the present invention, in step S1, the geometric parameters include thickness, width and camber parameters, and the scanning range for obtaining the geometric parameters of the steel strip covers three sampling points at the beginning, middle and end of the entire roll of steel strip to eliminate the influence of uneven distribution of residual stress.

[0016] According to one embodiment of the present invention, in step S1, the material property parameters are obtained by using a Rockwell hardness tester to uniformly select a preset number of sampling points on the surface of the steel strip for testing. Based on the hardness-strength correspondence curve, the Rockwell hardness value is converted into yield strength, tensile strength and elongation at break, and the arithmetic mean is taken as the representative value of the material property parameters.

[0017] According to one embodiment of the present invention, in step S3, the process of establishing the finite element simulation model includes sub-steps such as geometric model simplification, mesh generation, boundary condition setting, material model configuration, contact algorithm setting, and solver configuration.

[0018] According to one embodiment of the present invention, in step S4, a mapping relationship between shear quality characteristic parameters and process parameters is established. The mapping relationship between shear quality characteristic parameters and process parameters is established using a parametric analysis method. The process of establishing the mapping relationship is as follows: change the value of the process parameters, repeat steps S1 to S4 to obtain shear quality characteristic parameters corresponding to different combinations of process parameters, and establish a sample dataset between the process parameter space and the quality characteristic space.

[0019] According to one embodiment of the present invention, in step S5, the machine learning algorithm adopts a deep neural network. The neural network structure includes an input layer, multiple hidden layers and an output layer. The number of neurons in the input layer is the process parameter dimension, and the number of neurons in the output layer is four shear quality feature parameters.

[0020] According to one embodiment of the present invention, in step S6, the parameter range of the multi-parameter combination simulation calculation includes the shearing speed range, the tool gap range, the tool overlap range, and the tension range. The parameter combination search adopts the Latin hypercube sampling method, and the simulation calculation is performed in parallel to shorten the calculation time.

[0021] According to one embodiment of the present invention, in step S6, the optimal combination of process parameters for shearing quality is selected using a multi-objective optimization method. The four quality characteristic variables, namely burr height, collapse depth, cross-sectional roughness and dimensional deviation, are used as optimization objectives, and a weighted comprehensive evaluation function is constructed to sort the parameter combinations.

[0022] The weighted comprehensive evaluation function is:

[0023] ;

[0024] in, , , , These are the normalized values ​​for burr height, collapse depth, cross-sectional roughness, and dimensional deviation, respectively.

[0025] Weights of each quality characteristic variable , , , Determined based on the steel strip grade;

[0026] The normalization process uses the min-max normalization method, and the calculation formula is as follows: ;

[0027] in, It is a normalized value. These are the original values ​​of the quality characteristic variables. This represents the minimum value of the variable in the sample of parameter combinations. This represents the maximum value of the variable in the parameter combination sample.

[0028] According to one embodiment of the present invention, in step S7, the triggering condition for the process parameter adjustment command is that the proportion of the prediction result deviating from the threshold exceeds a preset deviation proportion, and the deviation proportion is calculated as follows: %,in For the first The deviation ratio of each quality characteristic variable, These are the predicted values ​​for quality characteristic variables. The threshold value is set for quality characteristic variables. When the deviation ratio of any quality characteristic variable exceeds the preset deviation ratio, it is determined that there is a risk of exceeding the standard for shearing quality under the current process parameter combination, and an automatic process parameter adjustment instruction is generated.

[0029] According to one embodiment of the present invention, in step S7, the process parameter adjustment instruction includes the parameter type to be adjusted, the parameter adjustment direction, and the parameter adjustment range. The parameter type to be adjusted includes shearing speed, tool clearance, tool overlap, and tension. The parameter adjustment direction is determined based on the deviation direction between the predicted value and the threshold. If the predicted burr height is higher than the threshold, the tool clearance needs to be increased or the shearing speed needs to be decreased. If the predicted collapse depth is greater than the threshold, the tool clearance needs to be decreased or the tension needs to be increased. The parameter adjustment range is determined based on the magnitude of the quality deviation and the parameter sensitivity, and an adaptive adjustment strategy is adopted. The larger the deviation, the larger the adjustment range, while limiting the single adjustment range to avoid parameter oscillation.

[0030] Compared with the prior art, the present invention has the following beneficial effects:

[0031] 1. High-precision quality control capability

[0032] Breaking through the bottleneck of experience dependence: By establishing a simulation calculation system based on digital twin model, the theoretical and quantitative analysis of the shearing process has been realized. The setting of process parameters no longer depends on the subjective experience of operators, the consistency of product quality has been significantly improved, and the batch-to-batch volatility has been controlled at an extremely low level.

[0033] Multi-parameter collaborative optimization: By systematically simulating and optimizing key process parameters such as shearing speed, tool clearance, tool overlap, and tension, process conflicts caused by single parameter adjustment are eliminated, and multi-parameter collaborative optimal control is achieved.

[0034] 2. Highly efficient production organization capabilities

[0035] Parameter configuration time is greatly reduced: By optimizing the parameter prediction function, the optimal combination of process parameters can be selected before production. The parameter debugging time is greatly reduced from the traditional method to within the preset time, and the efficiency of production preparation is significantly improved.

[0036] Real-time dynamic adjustment mechanism: The quality prediction model analyzes and processes real-time operating data and outputs prediction results. The system automatically triggers parameter adjustment commands based on prediction deviations, and can complete the fine-tuning of process parameters without stopping the machine, thus significantly improving the effective operating rate of the equipment.

[0037] 3. High stability and quality assurance capabilities

[0038] Effectively address material performance fluctuations: The digital twin model can sense changes in the properties of steel strip materials in real time and incorporate them into simulation calculations. The quality prediction model can automatically adjust parameters and recommend strategies based on changes in material properties, significantly enhancing its adaptability to material performance fluctuations.

[0039] Proactive quality prediction: The optimized parameter prediction function can predict the shearing quality results under specific process parameters before production, and select the optimal parameter combination for production accordingly, transforming quality defects from passive detection to proactive prevention. Attached Figure Description

[0040] Figure 1 This is a flowchart of a method for controlling the cross-cutting of steel strips used in the processing of advanced ultra-high strength steel plates and their coated plates according to an embodiment of the present invention;

[0041] Figure 2 This is a flowchart illustrating the logical flow of acquiring geometric parameters and material property parameters of steel strip and obtaining real-time working condition data in the steel strip cross-cutting control method according to an embodiment of the present invention.

[0042] Figure 3 This is a flowchart illustrating the logical flow of finite element simulation calculation and shear quality characteristic parameter extraction in the steel strip cross-cutting control method according to an embodiment of the present invention.

[0043] Figure 4 This is a schematic diagram of the core principle framework of the steel strip cross-cutting control method according to an embodiment of the present invention, which is based on the fusion of a digital twin model and a machine learning quality prediction model.

[0044] Figure 5 This is a schematic diagram of the multi-level interaction relationship and data flow of process parameter optimization prediction and dynamic adjustment control in the steel strip cross-cutting control method according to an embodiment of the present invention. Detailed Implementation

[0045] Example 1

[0046] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0047] like Figure 1 As shown, this invention provides a method for controlling the cross-cutting of steel strips used in the processing of advanced ultra-high strength steel plates and their coated plates. This method organically combines multiple technical aspects, including digital twin modeling, real-time working condition data acquisition, finite element simulation calculation, extraction of shearing quality characteristic parameters, construction of a quality prediction model, optimization parameter prediction, and dynamic parameter adjustment control. This achieves precise, quantitative, and predictable control of the steel strip cross-cutting process. This method is applicable to cross-cutting processing scenarios for various steel strip materials, such as advanced ultra-high strength steel plates, galvanized steel plates, and aluminized zinc steel plates, effectively ensuring shearing quality stability, improving production efficiency, and reducing process debugging costs.

[0048] Step 1: Establish a digital twin model of the steel strip cross section

[0049] In step 1 of this invention, establishing a digital twin model of the steel strip cross-section is the foundation and core starting point of the entire control method. For example... Figure 2 As shown, this digital twin model aims to construct a virtual mapping that is highly consistent with the physical steel strip cross-cutting production line in terms of geometry, materials, equipment, and processes, providing a solid data foundation and model support for subsequent simulation calculations, quality predictions, and parameter optimization.

[0050] In the steel strip geometric parameter acquisition stage, a laser scanner meeting the required accuracy is used to perform three-dimensional scanning measurements on the steel strip surface. The laser scanner works by emitting a modulated laser beam onto the surface of the steel strip being measured. By receiving the reflected light from the steel strip surface and calculating the beam's time of flight or phase difference, it accurately acquires the three-dimensional coordinate point cloud data of the steel strip surface. During the scanning process, the laser scanner's scanning angle covers the upper surface, lower surface, and side areas of the steel strip, ensuring that the acquired geometric morphology information is complete and accurate.

[0051] The acquisition strategy for the geometric parameters of the steel strip involves scanning three sampling points: the beginning, middle, and end of the entire coil. The beginning sampling point is located at the start of unwinding the steel strip coil, the middle sampling point is located in the middle region of the coil, and the end sampling point is located at the end of winding. The purpose of selecting these three locations is to eliminate the influence of uneven residual stress distribution generated during the winding and unwinding processes. Due to the radial pressure exerted by the drum during winding, the steel strip generates unevenly distributed residual stress along its length. The beginning and end are constrained by the drum for a shorter time, resulting in relatively sufficient residual stress release, while the middle is constrained for the longest time, leading to the most significant accumulation of residual stress. By acquiring geometric parameters at these three locations and averaging or weighting the data, the geometric distribution of the entire coil of steel strip can be effectively characterized.

[0052] Geometric feature parameters of the steel strip are extracted from the acquired 3D point cloud data, including steel strip thickness, steel strip width, and camber parameters. The steel strip thickness is calculated using the Euclidean distance between points on the upper and lower surfaces of the point cloud data, with a measurement accuracy controlled within ±0.01 mm. The steel strip width is calculated using the horizontal distance between points on the left and right sides of the point cloud data, with a measurement accuracy controlled within ±0.05 mm. The camber parameter is defined as the ratio of the maximum deviation of the steel strip side edge in the horizontal plane to the steel strip length. It is obtained by fitting a straight line to the points on the steel strip side edge and calculating the maximum vertical distance from each point to the fitted line. The formula for calculating the camber parameter is as follows:

[0053] %

[0054] in, Indicates the sickle curve value. Indicates the first side of the steel strip The horizontal coordinate values ​​of each sampling point This represents the lateral coordinates of the corresponding points on the fitted line. This indicates the length of the steel strip sampling section.

[0055] In the process of obtaining the material property parameters of steel strip, a Rockwell hardness tester is used to test the surface hardness of the steel strip to indirectly obtain the mechanical property parameters of the material. The testing principle of the Rockwell hardness tester is to use a diamond cone indenter or a steel ball indenter to press into the surface of the steel strip under a preset test force, and then measure the indentation depth and convert it into a Rockwell hardness value. During the test, a preset number of sampling points are evenly selected on the surface of the steel strip for testing. The distribution strategy of the sampling points is to evenly arrange no less than the preset number of sampling points in the width direction of the steel strip, and to arrange the sampling points at equal intervals in the length direction of the steel strip.

[0056] From the Rockwell hardness test results, based on the pre-established hardness-strength correlation curve, the Rockwell hardness values ​​are converted into yield strength, tensile strength, and elongation at break parameters. This correlation curve is established based on factors such as the steel's chemical composition and heat treatment state, and is calibrated and verified through standard tensile tests. The arithmetic mean of the test results from all sampling points is taken as the representative value of the material property parameters for that coil of steel strip. Simultaneously, the root mean square deviation (RMS) of the material property parameters is calculated to assess the material uniformity. When the RMS deviation exceeds a preset threshold, it indicates that the material properties of the coil of steel strip fluctuate significantly. The system will automatically mark this batch of steel strip and introduce a material property fluctuation correction factor in subsequent simulation calculations.

[0057] The yield strength parameter reflects the critical stress value at which the steel strip material begins to undergo plastic deformation. The yield strength of advanced ultra-high strength steel plates typically ranges from 780 MPa to 1500 MPa. The tensile strength parameter reflects the maximum stress value that the steel strip material can withstand before tensile fracture. The tensile strength of advanced ultra-high strength steel plates typically ranges from 900 MPa to 2000 MPa. The elongation at break parameter reflects the plastic deformation capacity of the steel strip material before tensile fracture. The elongation at break of advanced ultra-high strength steel plates typically ranges from 5% to 15%. For coated steel materials, additional parameters such as coating thickness and coating composition are required for thermodynamic coupling analysis in simulation calculations.

[0058] In the shearing tool parameter acquisition stage, a dedicated tool measuring device is used to accurately measure the geometric and wear parameters of the shearing tool. This device employs optical measurement principles, using a high-definition industrial camera to magnify and image the tool cutting edge, and extracting the tool's geometric feature parameters based on image processing algorithms. Specific tool parameters include tool clearance parameters, tool overlap parameters, tool hardness parameters, and cumulative usage time parameters. The tool clearance parameter is defined as the vertical distance between the lower cutting edge of the upper tool and the upper cutting edge of the lower tool; this parameter directly affects the stress distribution and cross-sectional quality in the shearing zone. The tool overlap parameter is defined as the vertical displacement of the upper tool above the plane of the lower tool; this parameter affects the deformation mode and crack propagation path in the shearing zone. The tool hardness parameter is obtained through a hardness tester; advanced shearing tools are made of cemented carbide or high-speed steel, with hardness values ​​typically ranging from HRC60 to HRC70. The cumulative usage time parameter records the cumulative actual shearing time since the last tool replacement.

[0059] The measurement of tool wear parameters employs an optical comparison method. Specifically, a high-resolution image of the current tool cutting edge is compared and analyzed with an image of a standard new tool cutting edge. An edge detection algorithm is used to extract the cutting edge contour, and the change in the cutting edge radius is calculated. When the cutting edge radius increases to a preset threshold, it indicates that the tool wear has reached a point where replacement is necessary, and the system automatically generates a tool replacement reminder. The preset cutting edge radius wear threshold is determined based on the type and thickness of the steel strip material. For advanced ultra-high strength steel plates, the cutting edge radius wear threshold is set to a preset value. When the cutting edge radius exceeds this threshold, the height of the shearing burr will increase significantly, affecting the shearing quality.

[0060] Based on the obtained geometric parameters, material property parameters, tool parameters, and process adjustment parameters, a digital twin model of steel strip cross-cutting is constructed. The data structure of the digital twin model adopts a hierarchical organizational architecture, including a geometry layer, a material layer, an equipment layer, a process layer, and a simulation layer. The geometry layer stores the 3D point cloud data of the steel strip and the results of feature parameter extraction. The material layer stores the constitutive model parameters, mechanical property parameters, and thermophysical parameters of the steel strip. The equipment layer stores the geometric model, motion parameters, and wear state information of the shearing tool. The process layer stores the set values ​​of process adjustment parameters such as shearing speed, tool clearance, tool overlap, and tension. The simulation layer stores the finite element mesh model, boundary condition settings, and solver configuration information.

[0061] The construction process of the digital twin model includes several sub-steps, such as geometric model import, material parameter mapping, tool model loading, process parameter configuration, mesh model generation, and simulation environment initialization. The geometric model import sub-step preprocesses the point cloud data acquired by laser scanning, including denoising, thinning, resampling, and surface fitting, to generate a continuous surface model for finite element analysis. The material parameter mapping sub-step retrieves the corresponding constitutive model parameters from the material database based on the steel strip material type. The constitutive model adopts an elastoplastic constitutive relation, considering the strain hardening and strain rate effects of the material. The tool model loading sub-step substitutes the geometric parameters obtained from tool measurements into the tool's 3D model, updating the tool's current position and orientation. The process parameter configuration sub-step sets the velocity curve, displacement curve, and force load curve during the shearing process. The mesh model generation sub-step uses professional meshing software to generate tetrahedral or hexahedral meshes for the geometric model. The simulation environment initialization sub-step sets the simulation solver type, time step, convergence criteria, and output options.

[0062] Step 2: Collect real-time operating data

[0063] In step 2 of this invention, as follows: Figure 2As shown, during the shearing production process, various sensor devices are used to collect shearing condition data in real time, constructing a real-time condition dataset covering multiple dimensions such as shearing force, vibration, and temperature. This provides dynamic data support for the updating and iteration of the shearing process simulation model and the real-time input of the quality prediction model.

[0064] Shear force signals are acquired using a shear force sensor, which is installed between the frame of the shearing equipment and the cutter holder. This sensor measures the force transmitted from the upper cutter to the lower cutter during the shearing process. The shear force sensor operates on the strain gauge principle. When shear force is applied to the sensor, the elastic element inside the sensor undergoes elastic deformation. The strain gauge attached to the surface of the elastic element changes its resistance accordingly. This change in resistance is converted into a voltage signal output via a bridge circuit. The sensor output signal is then amplified, filtered, and converted from analog to digital by a signal conditioning circuit, resulting in digitized shear force time-series data.

[0065] The sampling frequency of the shear force sensor is set to a preset first sampling frequency, which is determined based on the dynamic characteristics of the shearing process. The duration of the shearing process is typically in the range of milliseconds to seconds. The high-frequency components in the shear force signal are related to the fracture characteristics of the material, and the sampling frequency needs to meet the requirements of the Nyquist sampling theorem, i.e., the sampling frequency should not be less than twice the highest frequency component of the shear force signal. The typical range of the first sampling frequency can be appropriately adjusted according to the type and thickness of the steel strip in practical applications. For advanced ultra-high strength steel plates with a large thickness, the shear force amplitude is large and changes drastically, so the first sampling frequency is set to a higher value; for thinner coated plates, the shear force amplitude is small and changes relatively smoothly, so the first sampling frequency can be set to a moderate value.

[0066] Vibration signals are acquired using accelerometers, which are installed along the critical vibration transmission path of the shearing equipment, including locations such as the tool holder, tool shaft, transmission mechanism, and frame. The accelerometers operate on a piezoelectric or capacitive principle, detecting changes in charge or capacitance caused by vibration in the sensing element and converting the vibration acceleration into an electrical signal output. The sensor output signal is then processed by a signal conditioning circuit to convert it into digitized vibration acceleration time-series data.

[0067] The accelerometer's sampling frequency is set to a preset second sampling frequency, which needs to be able to capture the high-frequency vibration components during the shearing process. The vibration signal generated during shearing has a wide frequency range; the low-frequency components are related to the overall stiffness characteristics of the equipment, while the high-frequency components are related to the microscopic fracture process of the material. In practical applications, the typical value range of the second sampling frequency can be set to a value higher than the first sampling frequency to ensure that the time and frequency domain characteristics of the vibration signal can be completely acquired.

[0068] Temperature field distribution data were acquired using an infrared thermometer, which performs non-contact measurement of the surface temperature of the steel strip and cutting tool. The infrared thermometer operates based on Planck's blackbody radiation law, detecting the infrared radiation energy emitted by the surface of the object being measured and converting it into a temperature value. The measurement range covers the typical temperature range of the steel strip shearing region. Temperature measurement accuracy indicators include temperature resolution and measurement accuracy. Temperature resolution determines the sensitivity to temperature changes, while measurement accuracy determines the level of deviation between the measured temperature value and the true value.

[0069] The spatial resolution of the infrared thermometer is determined through the design of the optical system, with the smallest target size that can be resolved in the shearing area being a preset value. The temperature measurement point arrangement strategy involves placing multiple points along the length of the steel strip on both sides of the shear line, and several points along the width of the steel strip along the shear line, forming a two-dimensional temperature field monitoring array covering the shearing-affected area. Simultaneously, temperature measurement points are placed on the tool surface to monitor the tool's temperature rise during the shearing process. Excessive tool temperature rise will affect the tool's hardness and lifespan; therefore, a tool temperature threshold needs to be set in the monitoring system for alarm purposes.

[0070] The data acquisition time window is set from a specific time period before shearing to a specific time period after shearing. The purpose of this time window is to capture the complete physical response signal during the shearing process, including the preloading stage before shearing, the material deformation and fracture stage during shearing, and the material springback stage after shearing. The start and end times of the time window are determined based on the shearing trigger signal. When the shearing start signal is detected, the data acquisition system begins recording data and continues for a preset time length after the shearing completion signal.

[0071] The real-time operating condition dataset is stored in a time-series database format. Each record contains fields such as timestamp, sampling point number, sensor type, measured value, and data quality flag. The timestamp field uses a high-precision clock synchronization protocol to ensure time consistency of multi-sensor data. Time deviation is controlled within the time synchronization accuracy range. The data quality flag field indicates the validity status of the measured value. When a sensor malfunctions, a signal is lost, or the value exceeds the limit, this flag is set to an anomaly. During the data preprocessing stage, abnormal data will be removed or interpolated for correction.

[0072] Step 3: Establish a simulation model of the shearing process

[0073] like Figure 3As shown, in step 3 of this invention, a finite element simulation model is established based on the digital twin model constructed in step 1 and the real-time working condition data collected in step 2. The stress field distribution, strain field distribution, and temperature field distribution in the shear region are obtained through finite element simulation calculations, and the simulation results are output. The finite element simulation model is a concrete implementation of the digital twin model at the simulation calculation level, reproducing the physical phenomena such as material deformation, fracture, and heat transfer during the shearing process through numerical simulation methods.

[0074] The process of establishing a finite element simulation model includes core sub-steps such as geometric model simplification, mesh generation, boundary condition setting, material model configuration, contact algorithm setting, and solver configuration.

[0075] The geometric model simplification sub-step involves reasonably simplifying the complete steel strip cross-section geometric model based on the stress characteristics and analysis requirements of the shear region. The simplification strategy involves extracting the steel strip segment and tool segment within the shear influence region as the research objects of the simulation model. The steel strip portion outside the shear region is replaced with rigid constraints or elastic supports to reduce computational scale while ensuring simulation accuracy. The extent of the shear influence region is determined according to Saint-Venant's principle, typically taking a region several times the steel strip thickness on both sides of the shear line.

[0076] The meshing sub-step employs adaptive meshing technology to discretize the simplified geometric model. The core idea of ​​adaptive meshing is to automatically adjust the mesh density based on the gradient of the calculation results, refining the mesh in stress concentration and material fracture regions, and sparsening it in regions with gentle stress, thereby improving computational efficiency while maintaining accuracy. Tetrahedral meshes are used, with the mesh size in the shear region set to 0.1 mm. This smaller mesh size accurately captures the drastic stress-strain changes in the shear region. The mesh size in the boundary region is set to 0.5 mm, a larger value to reduce the number of elements in the boundary region. Tetrahedral or hexahedral elements are used for the mesh type, and quadratic or cubic interpolation is employed for the element order to improve the accuracy of displacement and stress calculations.

[0077] The specific implementation process of adaptive meshing technology is as follows: First, an initial simulation calculation is performed using a uniform coarse mesh. Based on the calculation results, the mesh error distribution is evaluated, and regions with large errors are identified as mesh refinement targets. Then, new nodes are inserted into the refinement target regions, and the mesh is regenerated. The positions of the new nodes are determined according to the error gradient direction. Next, a second simulation calculation is performed on the refined mesh, and the mesh error is re-evaluated. If the mesh error still exceeds a preset threshold, the above refinement process is repeated until the mesh error converges or the preset maximum refinement level is reached. The judgment criterion for adaptive mesh refinement is based on a stress gradient threshold or an energy error threshold (0.1%). Mesh refinement is triggered when the stress gradient or energy error in the region exceeds the corresponding threshold.

[0078] The boundary condition setting sub-step involves applying displacement and load boundary conditions to the finite element simulation model. Displacement boundary conditions include fixed constraints or displacement control constraints at both ends of the steel strip, and given motion trajectory constraints for the cutting tool. The constraint method at both ends of the steel strip is determined based on the actual clamping state. If tension control is used, a distributed load equivalent to the tension is applied; if displacement control is used, a given displacement time history curve is applied. The cutting tool's motion trajectory constraint is determined based on the shearing speed parameters and the tool clearance parameters. The upper tool descends vertically to the shearing position and stops, while the lower tool remains stationary. After shearing is completed, the upper tool moves upward to reset.

[0079] The load boundary conditions include material fracture loads and thermal loads in the shear region. Material fracture loads are introduced through the failure criteria of the material model. When the stress-strain state of an element satisfies the failure criteria, the element's stiffness matrix degenerates to simulate crack initiation and propagation. Thermal loads include plastic heat generated by plastic deformation during shearing, frictional heat generated by the friction between the tool and the steel strip, and environmental heat exchange. Thermal loads are applied by converting the calculated plastic and frictional dissipation work into thermal energy in a certain proportion, applying it to the model as a volumetric or surface heat source.

[0080] The material model configuration sub-step involves selecting the corresponding constitutive models and material parameters for the steel strip and tool materials. The steel strip material uses an elastoplastic constitutive model, considering both strain hardening and strain rate effects. The strain hardening effect is described using an isotropic hardening model or a kinematic hardening model; during plastic deformation, the yield stress increases with increasing cumulative plastic strain. The strain rate effect is described using a viscoplastic model; under high-speed deformation conditions, the yield stress increases with increasing strain rate. The tool material uses a rigid body model or a high-hardness elastoplastic model; the tool's deformation during shearing is very small, so it is approximated as a rigid body to simplify calculations.

[0081] The contact algorithm defines the contact relationships between the upper tool and the steel strip, and between the steel strip and the lower tool through contact pairing. The algorithm employs either a penalty function contact algorithm or a Lagrange multiplier contact algorithm. By defining the principal and slave contact surfaces and applying contact constraints, it coordinates the normal displacement and transmits the tangential friction force at the contact interface. The friction model uses either the Coulomb friction model or a modified Coulomb friction model, with the friction coefficient determined based on the steel strip material and tool material type.

[0082] The solver configuration substep selects the explicit dynamics algorithm as the solution method for the finite element simulation model. The explicit dynamics algorithm is based on an explicit time integral scheme and uses the central difference method to solve the equations of motion, making it suitable for solving highly nonlinear and dynamic response problems. The explicit algorithm eliminates the need for iterative solutions to large simultaneous equation systems, has a smaller computational load at each step, and exhibits good computational stability, making it particularly suitable for problems involving material fracture, contact friction, and large deformation.

[0083] The simulation time step is set to 1×10. -6 The time step 's' needs to satisfy the stability condition of the explicit algorithm, i.e., the time step must be less than the critical time step of the smallest element. The critical time step is determined based on the size, material density, and elastic modulus of the smallest element. The simulation time step is typically taken as 0.8 to 0.9 times the critical time step to ensure numerical stability. The time integration process of the explicit algorithm is as follows: , ,in For the quality matrix, For node acceleration, For external load vector, For internal force vectors, The hourglass damping force vector. For nodal displacement, For time step.

[0084] After the finite element simulation is completed, the calculation results data of the shear region are extracted as the simulation result data output. The output results include stress field distribution data, strain field distribution data, and temperature field distribution data. The format of the result data adopts the standard finite element result file format, including node numbers, node coordinates, and nodal values ​​or element values ​​of each field quantity. The data volume of the result data is determined according to the mesh size and the number of output field quantities, and a balance needs to be struck between data compression and storage and result integrity. The simulation result data is directly passed to step 4 for the extraction of shear mass characteristic parameters.

[0085] Step 4: Extract shear quality characteristic parameters

[0086] like Figure 3 As shown, in step 4 of this invention, the simulation results output in step 3 are analyzed and processed to extract the shear burr height parameter, shear collapse angle depth parameter, cross-sectional roughness parameter, and dimensional accuracy parameter as shear quality characteristic parameters, and a mapping relationship between the shear quality characteristic parameters and process parameters is established. The core objective of this step is to convert high-dimensional simulation results data into quantifiable and evaluable shear quality evaluation indicators, providing output variables for the subsequent establishment of a quality prediction model.

[0087] The extraction of the shear burr height parameter employs a vision inspection system, which consists of an industrial camera, a light source, and image processing software. The industrial camera is either a line-scan or area-scan camera, with a resolution sufficient to clearly distinguish the edge contour of the burr. A coaxial or low-angle light source is used to enhance the contrast between the burr and the substrate. The image processing software preprocesses the acquired burr area image, including grayscale transformation, filtering and denoising, and edge enhancement. Then, an edge detection algorithm is used to extract the burr contour boundary, and the maximum vertical distance of the burr contour relative to the steel strip shearing plane is calculated as the burr height value.

[0088] The image processing algorithm for burr height extraction includes five sub-steps: image acquisition, image preprocessing, edge detection, contour extraction, and height calculation. The image acquisition sub-step uses an industrial camera to acquire images of the sheared steel strip end face at a predetermined distance behind the shear line. The image preprocessing sub-step converts the original image to grayscale to eliminate color channel information, performs spatial filtering to eliminate random noise, and frequency domain filtering to enhance edge features. The edge detection sub-step uses the Canny edge detection algorithm or a sub-pixel edge detection algorithm to extract the edge points of the burr contour. The contour extraction sub-step connects the edge points into a continuous edge contour and performs fitting and smoothing processing on the contour. The height calculation sub-step converts the image pixel coordinates to world coordinates based on the camera calibration parameters and calculates the vertical distance of the highest point of the burr contour relative to the reference plane.

[0089] The depth parameter of the shear collapse angle was extracted using a contact profilometer. This device uses a probe to contact the sheared end face of the steel strip and scan along the end face profile to measure the two-dimensional profile curve of the end face. The measurement accuracy indicators of the contact profilometer include profile resolution and morphological accuracy. Profile resolution determines the sensitivity of the collapse angle feature detection, while morphological accuracy determines the deviation level between the measured value and the true value. During the measurement process, the probe moves from one side of the sheared end face to the other along the width direction of the steel strip, and the recorded curve of the probe's vertical displacement is the end face profile curve.

[0090] The specific method for extracting the collapse angle depth parameter from the end face contour curve is as follows: First, identify the feature point corresponding to the shear plane position on the contour curve. This feature point is the inflection point where the contour curve transitions from steep to gentle. Then, calculate the vertical distance between the shear plane position and the lowest point of the contour curve; this distance is the collapse angle depth value. Finally, select multiple measurement positions within the width of the steel strip for measurement, and take the average or maximum value of the measurement results as the final representative value of the collapse angle depth parameter. The measurement accuracy of the collapse angle depth is controlled within a preset range, and the measurement repeatability is better than the preset repeatability threshold.

[0091] The surface roughness parameters were extracted using a surface roughness meter, which acquires the surface roughness information of the steel strip shear section through stylus-type or optical measurement principles. The measured parameter is the Ra value, i.e., the arithmetic mean deviation of the profile. This parameter is defined as the arithmetic mean of the absolute values ​​of the profile deviations within the sampling length, and the calculation formula is:

[0092]

[0093] in, For sampling length, This represents the vertical offset of the profile curve relative to the baseline. The Ra value comprehensively reflects the roughness of the cross-sectional surface; a larger Ra value indicates a rougher cross-section, while a smaller Ra value indicates a smoother cross-section. The surface roughness tester outputs measurement results in micrometers.

[0094] Dimensional accuracy parameters are extracted using a laser velocimeter. This device measures the speed of the steel strip in the shearing zone using the laser Doppler principle and calculates the shearing dimensional parameters by combining this with the shearing synchronization signal. The measurement accuracy indicators of the laser velocimeter include speed resolution and position accuracy. Speed ​​resolution determines the accuracy of the shearing speed detection, while position accuracy determines the accuracy of the calculated shearing dimensions. During the shearing process, the laser velocimeter measures the feed and discharge speeds of the steel strip in real time, calculates the actual shearing length of the steel strip based on the cutter shearing frequency, and compares this with the set shearing length to obtain the dimensional deviation value.

[0095] The calculation method for dimensional accuracy parameters is as follows: ,in For shear length deviation, The actual shear length measured by the laser velocimeter. The target shearing length is set for the process parameters. Evaluation indicators for dimensional accuracy parameters include absolute deviation value and relative deviation rate. The relative deviation rate is the ratio of the absolute deviation value to the target length, expressed as a percentage.

[0096] The mapping relationship between shear quality characteristic parameters and process parameters is established using parametric analysis. The process involves changing the values ​​of key parameters such as shearing speed, tool clearance, tool overlap, and tension, and repeating steps 1 to 4 to obtain the shear quality characteristic parameters corresponding to different combinations of process parameters, thus establishing a sample dataset between the process parameter space and the quality characteristic space. This sample dataset records the results of each combination of process parameters and its corresponding quality characteristic parameters, visualized as a quality response surface or contour plot in the process parameter space. The mathematical form of the mapping relationship can be established using methods such as polynomial fitting, radial basis function interpolation, or neural network approximation. The specific method chosen depends on the degree of nonlinearity of the mapping relationship and the characteristics of the data distribution.

[0097] Step 5: Establish a quality prediction model

[0098] In step 5 of this invention, a sample dataset is constructed based on historical production data. A machine learning algorithm is then used to train the sample dataset, establishing a quality prediction model between process parameters and shear quality characteristic parameters. This quality prediction model can output corresponding shear quality prediction results based on a given combination of process parameters, enabling advance prediction of shear quality and optimized selection of process parameters.

[0099] The sample dataset was constructed based on process parameter records and quality inspection records accumulated during historical production. Historical production data sources include process parameter settings recorded by the production execution system and quality inspection results recorded by the quality inspection system. Process parameter records include shearing speed parameters, tool clearance parameters, tool overlap parameters, tension parameters, strip thickness parameters, strip width parameters, yield strength parameters, and tensile strength parameters. Quality inspection records include burr height detection values, corner depth detection values, surface roughness Ra values, and dimensional deviation values.

[0100] The preprocessing of the sample dataset includes three sub-steps: data cleaning, data transformation, and data partitioning. The data cleaning sub-step performs integrity checks and outlier handling on the raw data, removing records with severe data loss or abnormal test results. The data transformation sub-step normalizes process parameters and quality characteristics of different dimensions, scaling parameter values ​​to the [0,1] range to eliminate the impact of dimensional differences on model training. The data partitioning sub-step divides the sample dataset into training, validation, and test sets at a ratio of 80%, 10%, and 10%, respectively. Stratified sampling is used during partitioning to ensure that the distribution of quality characteristics in each set is consistent with the overall distribution.

[0101] The machine learning algorithm employs a deep neural network, whose structure includes an input layer, multiple hidden layers, and an output layer. The input layer has 9 neurons, corresponding to the following process parameters: strip thickness (mm), strip width (mm), yield strength (MPa), tensile strength (MPa), elongation at break (%), shearing speed (mm / s), tool clearance (mm), tool overlap (mm), and tension (kN). The output layer has 4 neurons, corresponding to burr height (mm), corner depth (mm), surface roughness Ra (μm), and dimensional deviation (mm), respectively. There are 3 hidden layers, with 64, 32, and 16 neurons in each layer. The output layer uses a linear activation function to adapt to the regression prediction task.

[0102] Activation Functions and Regularization: After each hidden layer, a batch normalization layer and a ReLU activation function are added sequentially. Batch normalization is used to accelerate convergence and alleviate gradient vanishing; its calculation formula is as follows:

[0103]

[0104] in, The input features of the batch normalization layer (i.e., the activation values ​​output by the previous layer). These are the output values ​​after batch normalization; and These are the mean and standard deviation of the current mini-batch data on this feature dimension, respectively. Let be the numerical stability constant, and take . ; and These are learnable scaling and offset factors.

[0105] The expression for the ReLU activation function is: ,in For input values, This is the output value. To prevent overfitting, a Dropout layer is introduced after the ReLU activation of each hidden layer, with the Dropout ratio set to 0.2 (i.e., randomly dropping 20% ​​of neurons during training). The network weights are initialized using the He initialization method, starting from a mean of 0 and a variance of... Sampling is performed from a normal distribution, where This is the number of input neurons in the current layer; the bias term is initialized to 0.

[0106] Training Strategy and Optimizer: The optimizer uses the Adam algorithm with an initial learning rate of 0.001, and the first moment is used to estimate the decay rate. Second-order moment estimation of attenuation rate Numerical stability term The loss function uses mean squared error (MSE), and its calculation formula is as follows:

[0107]

[0108] in, This is the loss value; This is the batch size (32 in this example). For the first The true quality characteristic value of each sample; For the model to the first The predicted values ​​for each sample are used. Mini-batch gradient descent is employed, with a batch size of 32. The learning rate decay strategy is to multiply the learning rate by 0.9 every 100 epochs. Early stopping is also used: training stops and the model parameters with the minimum validation loss are rolled back if the validation set loss does not decrease for 20 consecutive epochs. The maximum number of training iterations is set to 500.

[0109] Model Validation and Performance Metrics: After training, evaluate the model performance on the test set. Evaluation metrics include root mean square error (RMSE) and coefficient of determination (COP). For this embodiment, the acceptable thresholds are set as follows: burr height RMSE ≤ 0.02mm, collapse depth RMSE ≤ 0.04mm, cross-sectional roughness RMSE ≤ 0.3μm, and dimensional deviation RMSE ≤ 0.1mm. The model can only be used for production prediction when all indicators meet the requirements.

[0110] Model update mechanism: When the cumulative amount of newly added production data reaches 500 sets, the system automatically triggers an incremental model update. The incremental update employs a fine-tuning strategy: using the parameters of the previous version as initial values, the model is trained for 50 epochs with the new data at a small learning rate (0.0001), retaining the original network structure. The updated model needs to be re-evaluated on the latest validation set; if its performance is lower than the old version, it is reverted. This mechanism ensures that the quality prediction model can adapt to long-term changes such as fluctuations in material properties and tool wear.

[0111] Step 6: Predict and optimize parameters

[0112] like Figure 4 As shown, in step 6 of this invention, based on the quality prediction model established in step 5, multiple sets of parameter combinations are simulated and calculated for the range of process adjustment parameters. The shearing quality results corresponding to each set of parameter combinations are predicted, and the process parameter combination with the optimal shearing quality is selected as the recommended parameter. The optimized parameter prediction function realizes the quality prediction and optimization of process parameter combination schemes before production, transforming traditional post-production quality inspection into pre-production quality prevention.

[0113] Step 5 establishes a quality prediction model that uses process parameters as input and shear quality characteristic parameters as output, creating a mapping relationship between the process parameter space and the quality characteristic space. Step 6 then performs a reverse search, that is, within the given allowable range of process parameters, it searches for the optimal combination of process parameters that achieves the best shear quality through sampling and simulation calculations. This process embodies the integrated application of digital twin models and machine learning quality prediction models: the digital twin model provides a physical mechanism description of the shearing process, while the machine learning quality prediction model provides the ability to quickly predict the mapping relationship between process parameters and quality characteristics. The integration of the two enables intelligent optimization design of process parameters.

[0114] The parameter ranges for optimized parameter prediction include the shearing speed range, tool clearance range, tool overlap range, and tension range. The lower and upper limits of the shearing speed range are determined based on the mechanical properties of the shearing equipment and the material characteristics of the steel strip. The shearing speed range for advanced ultra-high strength steel plates is typically within a preset range. The lower and upper limits of the tool clearance range are determined based on the machining accuracy of the tools and the shearing quality requirements. Insufficient tool clearance may lead to tool collision interference, while excessive tool clearance may increase burr height. The lower and upper limits of the tool overlap range are determined based on the strength and life requirements of the tools. Insufficient tool overlap may lead to incomplete shearing, while excessive tool overlap may lead to accelerated tool wear. The lower and upper limits of the tension range are determined based on the material properties of the steel strip and the load-bearing capacity of the equipment. Insufficient tension may cause the steel strip to sag and become unstable, while excessive tension may cause tensile deformation of the steel strip. The tension range is typically within a preset range.

[0115] The parameter combination search employs the Latin hypercube sampling method, a multidimensional stratified sampling technique that can uniformly distribute sampling points within the parameter space while reducing the number of samples. The basic principle of Latin hypercube sampling is to divide each parameter dimension into several equally probable intervals, then randomly select a sample point from each interval, and finally randomly combine the sample points from each dimension to form a parameter combination. The advantage of this method is that the sampling points are uniformly distributed throughout the parameter space, avoiding the sample clustering problem that may occur with simple random sampling.

[0116] The sampling size is set to a preset number of sampling groups. Determining the sampling size requires a balance between computational accuracy and efficiency. Too few samples may result in insufficient coverage of the parameter space, making it impossible to accurately identify the optimal quality region; too many samples may lead to excessive computational costs, affecting the real-time performance of optimization predictions. A typical range of sampling size values ​​is a preset value, which can be adjusted based on the dimensionality of the parameter space and the complexity of the quality response surface. When the dimensionality of the parameter space is high or the nonlinearity of the quality response surface is high, the sampling size should be appropriately increased to ensure the reliability of the optimization results.

[0117] To shorten computation time, simulation calculations are performed in parallel, employing a distributed computing framework. This framework divides the parameter combination sample set into multiple subsets, with each computing node responsible for processing the simulation tasks within one subset. Computing nodes coordinate data and tasks via a high-speed network. After simulation completion, results are aggregated to the master node for quality evaluation and parameter ranking. The speedup of parallel computing is approximately linearly related to the number of computing nodes. However, as the number of nodes increases to a certain scale, the speedup tends to saturate, limited by the overhead of task partitioning and data transmission. The time consumption for parallel simulation execution is calculated as the single-task computation time divided by the number of computing nodes, plus the overhead of task scheduling and data transmission, significantly reducing computation time compared to serial computing.

[0118] The evaluation of the quality prediction results employs a multi-objective optimization method, using four quality characteristic variables—burr height, corner depth, surface roughness, and dimensional deviation—as optimization objectives, while also considering process parameter constraints. The solution strategy for multi-objective optimization involves constructing a weighted comprehensive evaluation function, integrating multiple optimization objectives into a single objective through weighted summation. The weighting coefficients are determined based on the relative importance of each quality characteristic variable and the quality control requirements. For the shearing process of advanced ultra-high strength steel plates, the weights of burr height and corner depth are typically set to higher values.

[0119] The weighted comprehensive evaluation function is:

[0120] ;

[0121] in, , , , These are the normalized values ​​for burr height, collapse depth, cross-sectional roughness, and dimensional deviation, respectively.

[0122] Weights of each quality characteristic variable , , , Determined based on steel strip grade:

[0123] For advanced ultra-high strength steel plates, the weight is set as follows: , , , ;

[0124] For ordinary steel strips with a yield strength ≤ 780 MPa, the weight is set as follows: , , , ;

[0125] The normalization process uses the min-max normalization method, and the calculation formula is as follows: ;

[0126] in, It is a normalized value. These are the original values ​​of the quality characteristic variables. This represents the minimum value of the variable in the sample of parameter combinations. This represents the maximum value of the variable in the parameter combination sample.

[0127] The target values ​​of each quality characteristic variable are preset according to the steel strip grade and customer requirements. The target values ​​for high-strength steel strips are relatively lenient, while the target values ​​for ordinary steel strips are relatively strict. The target values ​​are stored in the system database according to the steel strip grade and are used as the evaluation benchmark when performing multi-parameter combination simulation calculations in step S6.

[0128] The process for selecting the optimal combination of process parameters for shearing quality is as follows: First, calculate the predicted values ​​of the four quality characteristics corresponding to each parameter combination; then, calculate the comprehensive quality score of each parameter combination based on the weighted comprehensive evaluation function; finally, sort all parameter combinations in descending order of comprehensive quality score, and output the one or more parameter combinations with the highest comprehensive quality score as recommended parameters. The output of the recommended parameters includes the recommended setpoints for each process parameter, the expected predicted values ​​of quality characteristics, and the comprehensive score. The recommended parameters are directly passed to step 7 for parameter configuration of the cross-cutting equipment.

[0129] Step 7: Parameter Optimization and Dynamic Adjustment

[0130] In step 7 of this invention, the parameters of the cross-cutting equipment are configured according to the recommended parameters output in step 6. Real-time operating data is continuously collected during the shearing process and input into the quality prediction model for real-time prediction. When the prediction result deviates from the preset quality threshold, a process parameter adjustment command is automatically triggered, achieving dynamic optimization control of the shearing process. This dynamic adjustment control function realizes real-time monitoring and closed-loop control of the shearing quality, transforming quality control from static setting to dynamic adaptation. Figure 5 As shown, the closed-loop feedback mechanism of dynamic optimization control includes a quality prediction stage, a deviation assessment stage, a parameter adjustment stage, and an execution feedback stage, with each stage executing cyclically according to a fixed period.

[0131] The parameter configuration process for the cross-cutting equipment includes three sub-steps: parameter issuance, parameter verification, and parameter execution. The parameter issuance sub-step sends recommended parameters to the cross-cutting equipment control system via a network communication interface. The communication protocol uses either Industrial Ethernet or Fieldbus, and the communication cycle is controlled within a preset time range. The parameter verification sub-step checks the legality of the received parameter values, including parameter value range verification and parameter consistency verification, ensuring that the parameter values ​​are within the equipment's allowable range and match the current production conditions. The parameter execution sub-step involves the equipment control system writing the verified parameter values ​​into the corresponding control register, driving the actuators to operate according to the new parameter values.

[0132] The quality thresholds are determined based on the steel strip grade and customer requirements. The differences in quality thresholds for different grades of steel strip reflect the different quality control requirements. Advanced ultra-high strength steel plates, due to their high strength and high hardness, are more difficult to shear, and the burr height threshold and collapse depth threshold are usually set to relatively lenient values. Ordinary steel strips have relatively lower quality control requirements, and the burr height threshold and collapse depth threshold are usually set to relatively strict values.

[0133] The real-time prediction process is as follows: the data acquisition module continuously receives real-time operating condition data output by the sensors, including shear force signals, vibration signals, and temperature field distribution data; the data preprocessing module filters, calibrates, and normalizes the raw data, converting it into the input format required by the quality prediction model; the model inference module calls the quality prediction model to perform inference calculations on the preprocessed input data and outputs the predicted values ​​of four quality characteristic parameters; the result evaluation module compares the predicted values ​​with the corresponding quality thresholds to determine whether the predicted quality meets the requirements.

[0134] The trigger condition for the process parameter adjustment command is that the proportion of the predicted result deviating from the threshold exceeds the preset deviation proportion. The deviation proportion is calculated as follows: %,in For the first The deviation ratio of each quality characteristic variable, These are the predicted values ​​for quality characteristic variables. This refers to the threshold for quality characteristic variables. When the deviation ratio of any quality characteristic variable exceeds the preset deviation ratio, the system determines that there is a risk of exceeding the shearing quality limit under the current combination of process parameters and automatically generates a process parameter adjustment instruction.

[0135] The process parameter adjustment command includes the type of parameter to be adjusted, the direction of adjustment, and the adjustment range. The types of parameters to be adjusted include shearing speed, tool clearance, tool overlap, and tension. The adjustment direction is determined based on the deviation direction between the predicted value and the threshold. If the predicted burr height is higher than the threshold, the tool clearance needs to be increased or the shearing speed decreased; if the predicted burr depth is greater than the threshold, the tool clearance needs to be decreased or the tension increased. The adjustment range is determined based on the magnitude of the quality deviation and the parameter sensitivity, employing an adaptive adjustment strategy: the larger the deviation, the larger the adjustment range. Simultaneously, the adjustment range is limited to avoid parameter oscillation. The adjustment range per instance does not exceed 10% of the preset range, the preset deviation percentage threshold is set to 15%, and the control cycle is set to 1 second. After parameter adjustment, the parameter validity must be verified to avoid parameter oscillation. The adjustment command is sent to the cross-cutting equipment control system via a network communication interface. The control system executes the parameter adjustment operation upon receiving the command.

[0136] The closed-loop feedback mechanism of dynamic optimization control ensures the stability and robustness of the control effect. The control closed loop includes a quality prediction stage, a deviation assessment stage, a parameter adjustment stage, and an execution feedback stage, each stage executing cyclically according to a fixed period. The setting of the control period needs to balance control response speed and control stability; a control period that is too short may lead to control oscillation, while a control period that is too long may lead to control lag. A typical control period value is a preset value, which can be adjusted according to production rhythm and quality fluctuation characteristics.

[0137] This invention also includes establishing a steel strip cross-cutting database for storing and managing various data during the steel strip cross-cutting control process. The database records include steel strip batch information, equipment operating parameters, process parameter settings, real-time monitoring data, simulation calculation results, and quality inspection results. The database capacity is no less than a preset record quantity threshold, capable of storing production data records for at least a preset time period. The database supports categorized queries and statistical analysis by steel strip batch, equipment model, and process type, facilitating production managers and quality analysts to retrieve historical data, trace quality issues, and assess process stability.

[0138] The steel strip batch information record includes the steel strip grade, coil number, material performance parameters, geometric dimensional parameters, and supplier information. The equipment operation parameter record includes the equipment model, equipment status, tool model, cumulative tool usage time, and maintenance records. The process parameter setting record includes the set values ​​and adjustment history of each group of process parameters. The real-time monitoring data record includes raw sensor data, preprocessed data, and feature extraction results. The simulation calculation result record includes finite element simulation model parameters, boundary condition settings, and simulation result data. The quality inspection result record includes the detected values ​​of each quality characteristic variable, the judgment conclusion, and the analysis of the reasons for non-conformity.

[0139] The database query function supports both exact and fuzzy search modes. Exact searches return records that match exactly the specified criteria, such as querying the quality inspection results for a specific steel coil number. Fuzzy searches return records containing keywords, such as querying quality statistics for a certain type of material. The statistical analysis function supports quality trend analysis, parameter correlation analysis, and anomaly detection analysis. Quality trend analysis statistically analyzes the mean and fluctuation levels of quality indicators over different time periods to assess the trend of quality stability. Parameter correlation analysis uses regression or correlation analysis techniques to identify the quantitative relationship between process parameters and quality characteristics. Anomaly detection analysis, based on statistical process control or machine learning anomaly detection algorithms, identifies abnormal records where quality indicators deviate from the normal range and triggers alarms.

[0140] Example 2

[0141] To meet the needs of another application scenario, the present invention also provides an alternative implementation. This alternative implementation, while essentially retaining the technical solution of Embodiment 1, is adapted for specific steel grades or special working conditions. The alternative implementation is described in detail below.

[0142] For shearing processes involving coated steel sheets such as galvanized steel sheets and aluminized zinc steel sheets, the digital twin model construction process of this invention requires the addition of two extended modules: coating parameter acquisition and coating mechanical behavior modeling. The coating parameter acquisition module measures the coating thickness using an X-ray fluorescence spectrometer or an eddy current thickness gauge, with the measurement accuracy controlled within a preset range. The coating mechanical behavior modeling module employs an elastoplastic constitutive model to describe the mechanical behavior of the coating material, considering the interfacial bonding effect between the coating and the substrate.

[0143] The finite element simulation model needs to include interface contact elements between the coating and the substrate, as well as thermo-mechanical coupling analysis capabilities. The interface contact elements employ a Cohesive Zone model to describe the interfacial separation behavior between the coating and the substrate, with interface strength parameters determined through interfacial peeling tests. The thermo-mechanical coupling analysis needs to consider the impact of coating softening during shearing on the material behavior in the shear region; the temperature sensitivity parameters of the coating material are obtained through high-temperature mechanical tests.

[0144] During the extraction of quality characteristic parameters, the assessment of the shear quality of the coated plate requires the addition of a coating integrity evaluation index. This index involves acquiring images of the sheared end face using a visual inspection system to analyze the continuity and density of the coating in the sheared region. The coating integrity evaluation uses image processing algorithms to extract the coating edge contour and calculates the area ratio of the coating peeled off region as the evaluation index.

[0145] The foregoing has shown and described the basic principles and main features of the present invention. The scope of protection of the present invention is not limited to the above embodiments; any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. An advanced ultra-high strength sheet and a steel strip transverse cutting control method for processing plated sheet thereof, characterized by, Includes the following steps: S1. Based on the obtained geometric parameters, material property parameters, tool parameters, and process adjustment parameters of the steel strip, a digital twin model of steel strip cross-cutting is constructed as process parameters; S2. Collect shear force signals, vibration signals, and temperature field distribution data during the shearing production process to construct a real-time working condition dataset; S3. Based on the digital twin model and real-time working condition data, a finite element simulation model is established, and the stress field distribution, strain field distribution and temperature field distribution of the shear region are obtained through simulation calculation; S4. Extract the shear burr height, shear collapse angle depth, cross-sectional roughness, and dimensional accuracy as shear quality characteristic parameters; S5. Based on historical production data, a quality prediction model between process parameters and shear quality characteristic parameters is established using machine learning algorithms; S6. Based on the quality prediction model, perform simulation calculations on the range of process adjustment parameters using multiple parameter combinations, and select the optimal process parameter combination as the recommended parameters; S7. Configure the cross-cutting equipment according to the recommended parameters, continuously collect real-time working condition data during the shearing production process and input it into the quality prediction model for real-time prediction. When the prediction result deviates from the preset quality threshold, the process parameter adjustment command is automatically triggered to realize dynamic optimization control of the shearing process.

2. The method for controlling the cross-cutting of steel strip according to claim 1, characterized in that: In step S1, the geometric parameters include thickness, width, and camber parameters. The scanning range for obtaining the geometric parameters of the steel strip covers three sampling points at the beginning, middle, and end of the entire roll of steel strip to eliminate the influence of uneven distribution of residual stress.

3. The method for controlling the cross-cutting of steel strip according to claim 1, characterized in that: In step S1, the material property parameters are obtained by using a Rockwell hardness tester to uniformly select sampling points on the surface of the steel strip. Based on the hardness-strength correlation curve, the Rockwell hardness value is converted into yield strength, tensile strength and elongation at break, and the arithmetic mean is taken as the representative value of the material property parameters.

4. The method for controlling the cross-cutting of steel strip according to claim 1, characterized in that: In step S3, the process of establishing the finite element simulation model includes sub-steps such as geometric model simplification, mesh generation, boundary condition setting, material model configuration, contact algorithm setting, and solver configuration.

5. The method for controlling the cross-cutting of steel strip according to claim 1, characterized in that: In step S4, a mapping relationship between shear quality characteristic parameters and process parameters is established. The mapping relationship between shear quality characteristic parameters and process parameters is established using a parametric analysis method. The process of establishing the mapping relationship is as follows: change the value of the process parameters, repeat steps S1 to S4 to obtain shear quality characteristic parameters corresponding to different combinations of process parameters, and establish a sample dataset between the process parameter space and the quality characteristic space.

6. The method for controlling the cross-cutting of steel strip according to claim 1, characterized in that: In step S5, the machine learning algorithm uses a deep neural network. The neural network structure includes an input layer, multiple hidden layers, and an output layer. The number of neurons in the input layer is the process parameter dimension, and the number of neurons in the output layer is four shearing quality feature parameters.

7. The method for controlling the cross-cutting of steel strip according to claim 1, characterized in that: In step S6, the parameter ranges for simulation calculations using multiple parameter combinations include the shearing speed range, tool gap range, tool overlap range, and tension range. The parameter combination search uses the Latin hypercube sampling method, and the simulation calculations are performed in parallel to shorten the calculation time.

8. The method for controlling the cross-cutting of steel strip according to claim 1, characterized in that: In step S6, the optimal combination of process parameters for shearing quality is selected using a multi-objective optimization method. The four quality characteristic variables, namely burr height, collapse depth, cross-sectional roughness and dimensional deviation, are used as optimization objectives. A weighted comprehensive evaluation function is constructed to sort the parameter combinations. The weighted comprehensive evaluation function is: ； wherein, , , , are the normalized values of burr height, undercut depth, surface roughness and dimensional deviation, respectively. weight of each quality characteristic variable , , , determined according to the steel strip grade; The normalization processing adopts a min-max normalization method, and a calculation formula is: ; wherein is a normalized value, is an original value of the quality characteristic variable, is a minimum value of the variable in the parameter combination sample, is a maximum value of the variable in the parameter combination sample.

9. The method for controlling the cross-cutting of steel strip according to claim 1, characterized in that: In the step S7, the trigger condition of the process parameter adjustment instruction is that the proportion of the deviation of the prediction result from the threshold value exceeds a preset deviation proportion, and the calculation method of the deviation proportion is: %, wherein is the deviation proportion of the jth quality characteristic variable, is the prediction value of the quality characteristic variable, is the threshold value of the quality characteristic variable. When the deviation proportion of any quality characteristic variable exceeds the preset deviation proportion, it is determined that there is an over-standard risk of the shearing quality under the current process parameter combination, and the process parameter adjustment instruction is automatically generated.

10. The method for controlling the cross-cutting of steel strip according to claim 1, characterized in that: In step S7, the process parameter adjustment instruction includes the type of parameter to be adjusted, the direction of parameter adjustment, and the magnitude of parameter adjustment. The type of parameter to be adjusted includes shearing speed, tool clearance, tool overlap, and tension. The direction of parameter adjustment is determined based on the deviation direction between the predicted value and the threshold. If the predicted burr height is higher than the threshold, the tool clearance needs to be increased or the shearing speed needs to be decreased. If the predicted collapse depth is greater than the threshold, the tool clearance needs to be decreased or the tension needs to be increased. The magnitude of parameter adjustment is determined based on the magnitude of the quality deviation and the parameter sensitivity. An adaptive adjustment strategy is adopted, with a larger adjustment magnitude for larger deviations, while limiting the magnitude of a single adjustment to avoid parameter oscillation.

Images