Positioning and clamping system and method for laser cladding repair of large work rolls in steel plants

Abstract

This invention relates to the field of laser cladding repair technology, providing a positioning and clamping system and method for laser cladding repair of large work rolls in steel plants. A coupled prediction model of temperature, stress, and displacement fields is established. A partial least squares regression algorithm combined with a Kalman filter is used to achieve multi-step look-ahead prediction. Based on the predicted thermal deformation, an adaptive compensation control is implemented through a three-layer cascaded compensation mechanism consisting of a hydraulic drive, a piezoelectric ceramic actuator, and a laser head adjustment mechanism. The system uses time-domain, frequency-domain, and time-frequency domain joint analysis methods to extract vibration characteristics, identifies vibration types through support vector machine classification, and employs an active-passive hybrid suppression strategy for different vibration sources. This invention achieves high-precision positioning and stable clamping of large work rolls during laser cladding repair, significantly improving the thickness uniformity and surface quality of the cladding layer.

Description

Technical Field

[0001] This invention relates to the field of laser cladding technology, and specifically to a positioning and clamping system and method for repairing the surface of large work rolls in steel plants based on laser cladding. Background Technology

[0002] Large work rolls in steel mills are core components of steel rolling production lines. Operating under harsh conditions of high temperature, high pressure, and intense friction for extended periods, their surfaces are prone to defects such as wear, peeling, and cracks, directly impacting the quality of rolled products and production efficiency. Laser cladding technology, as an advanced surface repair method, uses a laser beam to melt alloy powder and coat it onto the work roll surface, forming a high-performance coating that is metallurgically bonded to the substrate. This effectively repairs surface defects and extends the service life of the work rolls. Compared to traditional repair methods such as welding and spraying, laser cladding offers significant advantages, including a smaller heat-affected zone, lower dilution rate, and higher coating quality, leading to its widespread application in work roll repair.

[0003] In the laser cladding repair of large work rolls, the work roll needs to rotate under the support of a positioning and clamping system, while the laser head moves axially to scan and clad. The main function of the positioning and clamping system is to accurately position and stably clamp the work roll, ensuring that the laser beam and the surface of the work roll always maintain the correct relative position, which is crucial for ensuring the uniformity of the cladding layer thickness and surface quality. However, the laser cladding process is a complex multi-physics coupling process. Local heating of the laser beam can cause significant thermal deformation of the work roll, and the rotation of the work roll and laser impact can cause system vibration. These factors can seriously affect the positioning accuracy and cladding quality. Existing positioning and clamping systems have problems in thermal deformation compensation. Traditional methods usually use passive response compensation, that is, compensation is only performed after thermal deformation is detected. This method has obvious lag and it is difficult to track the thermal deformation trend of the work roll in real time. Most existing compensation mechanisms use a single-level adjustment method, which is difficult to balance the dual requirements of compensation range and compensation accuracy. It lacks sufficient adjustment range for large thermal deformation and is difficult to achieve high-precision compensation for small deformation. Furthermore, existing methods lack accurate modeling of the coupling relationship between temperature, stress, and displacement fields, making it impossible to accurately predict the thermal deformation behavior of the work roll. This results in low compensation accuracy and slow response speed, severely affecting the relative positional accuracy between the laser beam and the work roll surface. During rotation, the work roll generates various types of vibrations, including unbalanced vibration, bearing failure vibration, laser shock vibration, and resonant vibration, each with different generation mechanisms and frequency characteristics. Traditional methods typically employ only passive vibration reduction measures, such as installing dampers or vibration isolation platforms, using the same suppression strategy for all types of vibration. This lacks specificity and results in limited vibration suppression effectiveness. Existing methods lack in-depth analysis and accurate identification of vibration signals, failing to distinguish between different types of vibration sources and thus unable to implement targeted suppression measures. The presence of vibration causes relative displacement between the laser beam and the work roll surface, making the molten pool unstable and severely affecting the surface quality and thickness uniformity of the cladding layer. The alignment accuracy between the work roll axis and the laser processing axis directly affects the cladding quality. However, during laser cladding, the work roll axis is prone to misalignment due to various factors such as thermal deformation, gravity, and support deformation. Traditional alignment methods typically rely solely on mechanical adjustment, using a servo motor to drive a support frame to move and adjust the position of the work roll. While this method offers high positioning accuracy, it is slow, and mechanical impacts can negatively affect the work roll. Existing methods lack an alignment mechanism that leverages the combined effects of multiple physical fields, failing to fully utilize the complementary advantages of different adjustment methods. This results in a lengthy alignment process and ultimately lower accuracy. Furthermore, existing axis deviation detection methods often employ a single measurement method, leading to limitations in measurement accuracy and reliability, and susceptibility to environmental interference and measurement noise.

[0004] Therefore, there is an urgent need to develop a new positioning and clamping system and method to improve the quality and efficiency of laser cladding repair, which is of great practical significance for promoting the advancement of large work roll repair technology. Summary of the Invention

[0005] To address the problems existing in the background art, the present invention provides a positioning and clamping method for repairing the surface of large work rolls in steel plants based on laser cladding, comprising the following steps:

[0006] S1. Establish a multi-physics field coupled prediction model: Arrange multiple measurement sections along the axial direction on the surface of the working roll. Each section is equipped with a temperature sensor, a strain sensor and a displacement sensor to establish a three-field coupled prediction model of temperature field, stress field and displacement field.

[0007] S2. Perform multi-step look-ahead prediction: Use partial least squares regression algorithm combined with Kalman filter to establish a combined prediction model, and predict the thermal deformation at future moments based on historical data of temperature, stress and displacement.

[0008] S3. Implement adaptive compensation control: Calculate the compensation control amount based on the predicted thermal deformation amount, and perform compensation through a three-layer cascaded compensation mechanism, including coarse compensation of hydraulically driven V-shaped bracket, fine compensation of piezoelectric ceramic actuator and ultra-precise compensation of laser head Z-axis adjustment mechanism.

[0009] S4. Perform vibration feature extraction: Collect vibration signals at key locations of the positioning and clamping system, and extract time-domain features, frequency-domain features, and time-frequency-domain features;

[0010] S5. Perform vibration source classification and suppression: Use support vector machine to classify and identify vibration features, and adopt an active-passive hybrid suppression strategy according to the identified vibration type.

[0011] S6. Perform axis deviation detection: Use three methods, laser scanning, machine vision and strain field detection, to detect the axis deviation of the work roll;

[0012] S7. Perform multi-field joint alignment: Based on the detected axis deviation, start the thermal gradient field alignment system, electromagnetic force field alignment system and mechanical force field alignment system, and use the sequential quadratic programming algorithm to coordinately optimize the control parameters of the three fields;

[0013] S8. Perform laser cladding process control: During the cladding process, perform multiphysics field coupling prediction and compensation, vibration suppression and dynamic centering maintenance in parallel.

[0014] Furthermore, S1 specifically includes:

[0015] S11. Configure the sensor array: Arrange the sensor array evenly along the axial direction on the surface of the work roll. Each measurement section is equipped with multiple infrared temperature sensors to collect temperature field data, multiple strain gauge sensors to collect stress field data, and multiple laser displacement sensors to collect displacement field data.

[0016] S12. Establish a coupled prediction model: Based on the collected temperature field data... Stress field data and displacement field data Establish a three-field coupling relationship, including the temperature field. The stress field is determined by laser power, scanning speed, and substrate temperature. From temperature field The coefficient of thermal expansion and the elastic modulus are determined, and the displacement field is... From stress field The moment of inertia and length of the working roller are determined.

[0017] Furthermore, S2 specifically includes:

[0018] S21. Extract latent variables: The temperature of each measurement section... ,stress Displacement Historical data constitutes the input matrix The corresponding thermal deformation amounts are used to form the output matrix. Extracted by partial least squares regression One potential variable;

[0019] S22. Establishing the state-space model: Constructing the state equations of the Kalman filter. and observation equations ;in for The state vector at any given time includes the temperature, stress, and displacement values ​​of each measured section; for The input vector at each moment includes laser power, scanning speed, and rotational angular velocity; This is the state transition matrix; The input matrix; This is process noise; for The observation vector at time; The observation matrix; To observe noise;

[0020] S23. Perform multi-step prediction: based on the current time... state vector and input vector The future is calculated recursively using the state equation. Thermal deformation at step time ,in For prediction Thermal deformation at any given time (mm); To predict the number of steps.

[0021] Furthermore, S3 specifically includes:

[0022] S31. Calculate the compensation amount: Based on the predicted thermal deformation amount... Through the compensation control law Calculate radial compensation displacement ;in Radial compensation displacement (mm); For proportional gain; For prediction Thermal deformation at any given time (mm); This is the differential gain; For prediction Thermal deformation at any given time (mm);

[0023] S32. Adjust control gain: Based on the predicted thermal deformation. The magnitude and rate of change are adaptively adjusted by fuzzy rules to adjust the proportional gain. and differential gain ;

[0024] S33. Implement three-tier compensation: adjust the compensation amount. The compensation mechanism is distributed to three layers. It performs millimeter-level coarse compensation by driving the V-shaped bracket through the hydraulic system, micron-level fine compensation by the piezoelectric ceramic actuator, and submicron-level ultra-fine compensation by the laser head Z-axis adjustment mechanism.

[0025] Furthermore, S4 specifically includes:

[0026] S41. Vibration signal acquisition: Install acceleration sensors at the chuck, bearing housing, and V-bracket positions to acquire vibration acceleration signals. ;in for Vibration acceleration value at time (m / s) 2 ); Time (s);

[0027] S42. Extracting temporal features: Calculating the root mean square value Calculate the peak factor ;in The root mean square value (m / s) 2 ); This represents the number of sampling points; For the first Vibration acceleration values ​​at each sampling point (m / s²) 2 ); This is the summation operator; This is the square root operator; Peak factor; The maximum value of the vibration signal (m / s) 2 );

[0028] S43. Extracting frequency domain features: For vibration signals... The spectrum is obtained by performing a fast Fourier transform. Identify main frequency Calculate the spectral entropy ;in For frequency domain spectrum; Frequency (Hz); The main frequency is the frequency (Hz) corresponding to the maximum amplitude of the spectrum. For spectral entropy; For frequency Normalized energy probability at the location; For logarithmic operators;

[0029] S44. Extracting time-frequency domain features: For vibration signals... conduct Layer wavelet packet decomposition, calculating the th Layer energy ;in The number of wavelet packet decomposition layers; For the first Layer energy characteristics; For the first Layer Wavelet coefficients;

[0030] S45. Constructing Feature Vectors: Combining time-domain features, frequency-domain features, and time-frequency-domain features to form feature vectors. Used for classifying vibration sources. Among them, RMS is the root mean square value, which reflects the energy of the vibration signal. The larger the value, the more intense the vibration. CF is the peak factor, which reflects the waveform characteristics of the vibration signal. It is dimensionless and distinguishes different types of vibration. The peak factor of periodic vibration is small, while the peak factor of impact vibration is large. The dominant frequency is the frequency corresponding to the maximum amplitude of the spectrum, reflecting the main frequency components of the vibration. Different vibration sources have specific frequency characteristics.

[0031] Furthermore, S5 specifically includes:

[0032] S51. Classifying vibration sources: This involves classifying the feature vectors... Input the support vector machine classifier to identify the vibration type as unbalanced vibration, bearing failure vibration, laser shock vibration, or resonant vibration.

[0033] S52. Suppressing Unbalanced Vibration: Unbalanced vibration is passively suppressed by adjusting the position of the counterweight, and actively suppressed by adjusting the speed of the servo motor. The speed compensation algorithm is as follows: ;in for Angular velocity (rad / s) after compensation at any moment; To set the angular velocity (rad / s); This is the unbalanced compensation coefficient; The vibration amplitude (m / s) 2 ); It is a sine function; The rotational angular velocity (rad / s); The phase angle is in rad.

[0034] S53. Suppressing Bearing Vibration: For bearing vibration, passive suppression is achieved by adding a viscoelastic damper, and active suppression is achieved by adjusting the damping coefficient using a magnetorheological damper. Damping force... ;in for Damping force (N) at time t; For the current The varying damping coefficient (N·s / m); The control current (A) of the magnetorheological damper; for Vibration velocity at any given moment (m / s);

[0035] S54. Suppressing Laser Shock Vibration: For laser shock vibration, passive suppression is achieved through an air spring vibration isolation platform, and active suppression is achieved by generating a reverse force through a piezoelectric ceramic actuator. The control law is as follows: ;in for The actuator output force (N) at all times; The transfer function of the adaptive filter; For the Laplace operator; for The disturbance force (N) at any given time.

[0036] Furthermore, S6 specifically includes:

[0037] S61. Perform laser scanning inspection: Install laser profile scanners at both ends of the work roll to collect the profile point set at the roll end during the rotation of the work roll. The coordinates of the circle center were fitted using the least squares method. Calculate the eccentricity and eccentricity ;in For the first Coordinates of each contour point (mm); Number of data collection points; The center coordinates (mm) of the fitted circle; The eccentricity (mm) is measured by laser scanning. Reference axis coordinates (mm); The eccentricity angle (rad) is measured by laser scanning. It is the arctangent function;

[0038] S62. Perform machine vision inspection: Capture images of the roller end face using an industrial camera, identify the center coordinates of the circle using edge detection and circle recognition algorithms, and calculate the eccentricity. and eccentricity ;in The eccentricity (mm) measured by machine vision method; The eccentricity angle (rad) is measured by machine vision.

[0039] S63. Perform strain field testing: Collect strain distribution data at the V-shaped bracket contact point and calculate the eccentric load index. Determine whether eccentricity exists; among which This is the off-center load index; This represents the maximum strain value. This is the minimum strain value;

[0040] S64. Fusion Detection Results: The detection results from laser scanning, machine vision, and strain field methods are weighted and fused to obtain the final eccentricity. and eccentricity ;in The eccentricity after fusion (mm); The weights for the laser scanning method; The weights for machine vision methods; The eccentricity angle (rad) after fusion.

[0041] Furthermore, S7 and S8 specifically include:

[0042] S71. Start thermal gradient alignment: The annular electric heaters at both ends of the work roll heat the rollers in sections according to the eccentric direction, generating a temperature difference. thermally induced deformation ;in The value represents the amount of thermally induced deformation (mm). The coefficient of linear expansion (1 / ℃); The length of the working roll is in mm. Temperature difference (°C); The diameter of the working roller (mm);

[0043] S72. Activate electromagnetic field alignment: Electromagnetic attraction is generated through the electromagnet array on both sides of the V-shaped bracket. ;in Electromagnetic attraction (N); This refers to the number of coil turns. Current (A); The permeability of free space (H / m); Relative permeability; magnetic pole area (m 2 ); The air gap is in meters (m).

[0044] S73. Start the mechanical field centering: The position of the V-shaped bracket is adjusted by driving the V-shaped bracket through a servo electric ball screw, and the position closed-loop control is adopted. ;in for Control quantity at any given time (mm); For proportional gain; for The eccentricity at any given time (mm); This is the integral gain; This is the differential gain; for The eccentricity at any given time (mm);

[0045] S74. Perform multi-field collaborative optimization: Establish the optimization objective function. The optimal heating power is solved using a sequential quadratic programming algorithm. Electromagnetic current and mechanical displacement ;in To optimize the objective function; This is the eccentricity weighting coefficient; Eccentricity (mm); This is the thermal field weighting coefficient; Heating power (W); This is the electromagnetic field weighting coefficient; For mechanical field weighting coefficients;

[0046] In S8, the following are performed simultaneously during the laser cladding process: predicting thermal deformation through a multi-physics field coupling prediction model and compensating in real time through a three-layer compensation mechanism; identifying vibration types through a vibration analysis model and suppressing vibration through an active-passive hybrid approach; and monitoring axis deviation through a multi-modal detection system and maintaining alignment accuracy through a multi-field joint system.

[0047] This invention also provides a positioning and clamping system for repairing the surface of large work rolls in steel plants based on laser cladding, comprising:

[0048] The sensing system includes an array of temperature sensors, a strain sensor array, and a displacement sensor array arranged along the axial direction of the work roll to collect temperature, stress, and displacement field data; it also includes vibration sensors installed at key locations in the positioning and clamping system to collect vibration signals; and a laser profile scanner and an industrial camera installed at both ends of the work roll to detect axial deviation. The temperature sensor array, strain sensor array, displacement sensor array, and vibration sensor are connected to the control system via a data acquisition card, and the laser profile scanner and industrial camera are connected to the control system via a communication interface.

[0049] The execution system includes a three-layer cascaded compensation mechanism for thermal deformation compensation, an active-passive hybrid vibration damping mechanism for vibration suppression, and a multi-field joint alignment mechanism for alignment calibration. The three-layer cascaded compensation mechanism includes a hydraulically driven V-shaped bracket, a piezoelectric ceramic actuator mounted on the V-shaped bracket, and a Z-axis adjustment mechanism connected to the laser head. The active-passive hybrid vibration damping mechanism includes a viscoelastic damper and an air spring as passive vibration damping components, and a magnetorheological damper and a piezoelectric ceramic actuator array as active vibration damping components. The multi-field joint alignment mechanism includes annular electric heaters inside the chucks at both ends of the work roll, an electromagnet array on both sides of the V-shaped bracket, and a servo-driven electric ball screw that drives the V-shaped bracket. The hydraulic drive system, piezoelectric ceramic actuator, Z-axis adjustment mechanism, magnetorheological damper, piezoelectric ceramic actuator array, annular electric heater, electromagnet array, and servo-driven electric ball screw are all connected to the control system via actuators.

[0050] The control system includes a data acquisition and processing module, a predictive compensation control module, a vibration analysis and suppression control module, and a centering optimization control module. The data acquisition and processing module receives sensor data through a data acquisition card, performs preprocessing, and then transmits the data to the predictive compensation control module, the vibration analysis and suppression control module, and the centering optimization control module. The predictive compensation control module includes a partial least squares regression unit, a Kalman filter unit, and a compensation control unit. It establishes a multi-physics coupled prediction model, performs multi-step look-ahead prediction, and generates compensation control commands to be sent to the three-layer cascaded compensation mechanism. The vibration analysis and suppression control module includes a feature extraction unit, a support vector machine classification unit, and a vibration control unit. It extracts vibration features, identifies vibration types, and generates suppression control commands to be sent to the active-passive hybrid vibration reduction mechanism. The centering optimization control module includes a deviation detection unit, a sequential quadratic programming optimization unit, and a multi-field control unit. It integrates multi-modal detection results, optimizes multi-field control parameters, and generates centering control commands to be sent to the multi-field joint centering mechanism.

[0051] 10. The system according to claim 9, characterized in that the temperature sensor array includes an infrared temperature sensor and a thermal imaging camera; the infrared temperature sensor is divided into multiple measurement sections along the working roller axis, and the multiple infrared temperature sensors in each measurement section are evenly distributed circumferentially; the strain sensor array includes strain gauge sensors and fiber optic strain sensors, and the strain gauge sensors and infrared temperature sensors are arranged in a staggered manner; the displacement sensor array includes laser displacement sensors, which are arranged opposite each other in each measurement section; the vibration sensor includes a triaxial acceleration sensor, which is respectively mounted on the chuck, bearing seat and V-shaped bracket;

[0052] In the three-layer cascaded compensation mechanism, the hydraulically driven V-shaped bracket is connected to the fixed column via a hydraulic cylinder, which is controlled by a hydraulic control valve; the piezoelectric ceramic actuator is installed at the contact position between the V-shaped bracket and the work roller, with one end of the piezoelectric ceramic actuator fixedly connected to the V-shaped bracket and the other end in contact with the surface of the work roller; the Z-axis adjustment mechanism is connected to the laser head mounting bracket via a guide rail, and the drive end of the Z-axis adjustment mechanism is connected to the output shaft of the voice coil motor; the control commands for the three-layer compensation mechanism are output from the compensation control unit of the predictive compensation control module.

[0053] In the active-passive hybrid vibration reduction mechanism, a viscoelastic damper is installed between the bearing housing and the foundation; an air spring is installed between the V-shaped bracket support base and the foundation; a magnetorheological damper is installed between the bearing housing and the foundation, arranged in parallel with the viscoelastic damper, and the current input terminal of the magnetorheological damper is connected to the PWM controller; a piezoelectric ceramic actuator array is installed on the V-shaped bracket support base, and the drive voltage input terminal of the piezoelectric ceramic actuator is connected to the piezoelectric driver; the PWM controller and the piezoelectric driver receive control signals output from the vibration control unit of the vibration analysis and suppression control module;

[0054] In the multi-field joint centering mechanism, an annular electric heater is embedded in the inner wall of the chucks at both ends of the work roll. The annular electric heater is divided into multiple independent control zones, each of which is independently powered by a heating controller. An electromagnet array is mounted on the brackets on both sides of the V-shaped bracket, with the magnetic pole faces of the electromagnets facing the surface of the work roll. The coils of the electromagnets are powered by an electromagnetic driver. The nut end of the servo electric ball screw is connected to the movable column, and the screw end is connected to the output shaft of the servo motor. The movable column is connected to the fixed base via a guide rail. The heating controller, electromagnetic driver, and servo motor driver receive control signals output from the multi-field control unit of the centering optimization control module.

[0055] The hardware architecture of the control system includes an industrial computer, a motion control card, a data acquisition card, and a fieldbus. The industrial computer runs a real-time operating system and connects to the motion control card and the data acquisition card through an expansion interface. The motion control card connects to hydraulic control valve drivers, piezoelectric drivers, voice coil motor drivers, PWM controllers, heating controllers, electromagnetic drivers, and servo motor drivers. The data acquisition card connects to temperature sensors, strain sensors, displacement sensors, and vibration sensors. The laser profile scanner and industrial camera communicate with the industrial computer through the fieldbus.

[0056] The beneficial effects achieved by this invention are as follows:

[0057] First, this invention achieves high-precision positioning and stable clamping of large working rollers during laser cladding repair by establishing a multi-physics field coupled prediction model, employing frequency-domain and time-domain joint vibration analysis technology, and implementing a multi-field joint alignment calibration mechanism. This invention organically combines multiple technologies, including multi-physics field coupled prediction, multi-step look-ahead prediction, adaptive compensation control, intelligent classification and suppression of vibration sources, multi-modal deviation detection, and multi-field collaborative optimization alignment, forming a complete intelligent adaptive control system. The system innovation of this invention lies in constructing a three-in-one architecture of sensing, execution, and control systems. Comprehensive state monitoring is achieved through the integrated configuration of temperature sensor arrays, strain sensor arrays, and displacement sensor arrays; precise control is achieved through a three-layer cascaded compensation mechanism, an active-passive hybrid vibration reduction mechanism, and a multi-field joint alignment mechanism; and intelligent decision-making is achieved through predictive compensation control modules, vibration analysis and suppression control modules, and alignment optimization control modules. Compared with traditional methods, this invention achieves significant improvements in several key performance indicators, such as positioning accuracy, thermal deformation compensation accuracy, vibration suppression effect, alignment accuracy, and cladding layer quality.

[0058] Secondly, this invention employs a partial least squares regression algorithm combined with a Kalman filter to establish a combined prediction model. Based on historical data of the temperature, stress, and displacement fields, it predicts the thermal deformation at future moments, achieving a fundamental shift from passive response compensation to active predictive compensation. By establishing a coupled prediction model of the temperature, stress, and displacement fields, this invention accurately describes the complete physical process from laser thermal input to the final deformation of the working roll, providing a theoretical basis for accurate prediction. A three-layer cascaded compensation mechanism organically combines millimeter-level coarse compensation from the hydraulically driven V-shaped bracket, micrometer-level fine compensation from the piezoelectric ceramic actuator, and submicrometer-level ultra-precise compensation from the laser head Z-axis adjustment mechanism, achieving full-scale accurate compensation from millimeter to submicrometer levels. This multi-level compensation strategy ensures a sufficient compensation range to cope with large thermal deformations while achieving extremely high final compensation accuracy. Furthermore, this invention uses fuzzy rules to adaptively adjust the proportional and differential gains based on the magnitude and rate of change of the predicted thermal deformation, enabling the compensation control law to adapt to the complex and variable working conditions during laser cladding. By combining multi-step look-ahead prediction with adaptive compensation control, this invention significantly improves positioning accuracy and thermal deformation compensation accuracy, while greatly shortening the compensation response time, ensuring that the laser beam and the surface of the work roll always maintain the correct relative positional relationship.

[0059] Third, this invention employs time-domain, frequency-domain, and time-frequency domain combined analysis methods to extract vibration features, constructing a comprehensive feature vector including root mean square value, peak factor, dominant frequency, spectral entropy, and wavelet packet energy at each layer. By intelligently classifying the feature vector using a support vector machine classifier, this invention can accurately identify different types of vibration, such as unbalanced vibration, bearing fault vibration, laser shock vibration, and resonant vibration, providing a scientific basis for selecting targeted suppression strategies. This invention adopts a hybrid active-passive suppression strategy. The viscoelastic damper and air spring in the passive damping section provide basic broadband vibration suppression capabilities, while the magnetorheological damper in the active damping section achieves precise suppression of specific vibration modes by adjusting the damping coefficient in real time and generating a reverse control force through a piezoelectric ceramic actuator. This combined active and passive approach fully leverages the general effectiveness of passive vibration isolation for broadband vibrations and the targeted suppression capability of active control for specific frequency vibrations, achieving effective control of vibrations across the entire frequency band. Effective vibration suppression directly improves the stability of the laser cladding process, keeping the molten pool stable, thereby significantly reducing vibration amplitude and surface roughness, and improving the surface quality of the cladding layer.

[0060] Fourth, this invention proposes a multi-field joint alignment method, comprehensively utilizing the effects of three physical fields—thermal gradient field, electromagnetic field, and mechanical field—to achieve rapid and high-precision alignment calibration. Thermal gradient alignment generates a temperature difference by selectively heating specific locations on the work roll, using the thermal expansion effect to drive the axis movement; its effect is gentle and will not cause impact or damage to the work roll. Electromagnetic field alignment achieves non-contact adjustment of the axis position by coordinating the current of each electromagnet at different positions on the work roll, generating different attractive forces; its fast response speed enables rapid position adjustment. Mechanical field alignment uses a servo-driven electric ball screw to drive a V-shaped bracket for position adjustment, providing final precise positioning. This invention employs a sequential quadratic programming algorithm to globally and collaboratively optimize the control parameters of the three fields, minimizing alignment time and energy consumption while meeting alignment accuracy requirements, achieving complementary advantages of the three physical field characteristics. This invention also integrates laser scanning, machine vision, and strain field detection methods to detect work roll axis deviation. Through multi-source information fusion, it effectively reduces the random errors of single measurement methods and improves the accuracy and robustness of eccentricity detection. The multi-field joint centering method gives the centering process significant advantages such as high speed, high accuracy, and smooth adjustment, which significantly improves centering accuracy and speeds up centering.

[0061] Fifth, the sensing system, execution system, and control system of this invention work in a unified and coordinated manner, achieving comprehensive monitoring and control of the working roll's state. The sensing system, through the integrated configuration of temperature sensor arrays, strain sensor arrays, displacement sensor arrays, vibration sensors, a laser profile scanner, and an industrial camera, achieves comprehensive real-time monitoring of the temperature field, stress field, displacement field, vibration signals, and axis deviation, providing a complete and accurate data foundation for multi-physics coupling prediction, vibration analysis, and alignment calibration. The execution system, through the coordinated operation of a three-layer cascaded compensation mechanism, an active-passive hybrid vibration damping mechanism, and a multi-field joint alignment mechanism, achieves precise compensation for thermal deformation, effective suppression of vibration, and rapid calibration of axis deviation. The control system, through intelligent decision-making by the predictive compensation control module, vibration analysis and suppression control module, and alignment optimization control module, achieves a complete closed loop from data acquisition, state prediction, feature recognition to control output. During laser cladding, multi-physics coupling prediction and compensation, vibration analysis and suppression, and multi-field joint alignment technologies are executed simultaneously in parallel, forming an intelligent adaptive control system capable of coping with various complex disturbances during laser cladding. This systematic technological innovation ensures the stability of the repair process and the consistency of repair quality, significantly improving the uniformity of the cladding layer thickness and surface quality. It effectively enhances the overall quality and efficiency of laser cladding repair, and has significant application value and promising prospects for promotion. Attached Figure Description

[0062] Figure 1This is a comparison chart of the positioning accuracy and centering accuracy between Examples 1-3 and Comparative Examples 1-3. Figure 1 Sub-image (a) shows a comparison of positioning accuracy; sub-image (b) shows a comparison of centering accuracy.

[0063] Figure 2 This is a comparison chart of Examples 1-3 and Comparative Examples 1-3 regarding the accuracy of thermal deformation compensation and the compensation response time. Figure 2 Subplot (a) shows a comparison of thermal deformation compensation accuracy; subplot (b) shows a comparison of compensation response time.

[0064] Figure 3 This is a comparison chart of Examples 1-3 and Comparative Examples 1-3 regarding vibration amplitude and surface roughness. Figure 3 Subplot (a) shows a comparison of vibration amplitudes; subplot (b) shows a comparison of surface roughness.

[0065] Figure 4 This is a comprehensive comparison chart of the cladding layer quality indicators between Examples 1-3 and Comparative Examples 1-3. Figure 4 Subplot (a) shows a comparison of the uniformity of cladding layer thickness; subplot (b) shows a comparison of surface roughness.

[0066] Figure 5 These are normalized box plots of various performance indicators for the examples and comparative examples.

[0067] Figure 6 This is a comparison chart of the vibration suppression effects between Example 1 and Comparative Example 2. Figure 6 Subplot (a) shows a comparison of the time-domain waveforms of the vibration signals; subplot (b) shows a comparison of the frequency spectra of the vibration signals.

[0068] Figure 7 This is a comparison graph of the dynamic response curves of Example 1 and Comparative Example 3 regarding the centering process.

[0069] Figure 8 This is a comparison diagram of the cladding layer thickness distribution between Example 1 and Comparative Example 1. Figure 8 Subplot (a) shows the cladding layer thickness distribution of Example 1; subplot (b) shows the cladding layer thickness distribution of Comparative Example 1.

[0070] Figure 9 This is a diagram of the positioning and clamping system architecture for repairing the surface of large work rolls in steel plants based on laser cladding, as described in this invention.

[0071] Figure 10 This is a flowchart of the positioning and clamping method for repairing the surface of large work rolls in steel plants based on laser cladding, according to the present invention. Detailed Implementation

[0072] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings. In addition, the forms of the various structures described in the following embodiments are merely illustrative. The present invention is not limited to the structures described in the following embodiments. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0073] Reference Figure 9 This invention provides a positioning and clamping method for repairing the surface of large work rolls in steel plants based on laser cladding, comprising three main parts: a sensing system, an execution system, and a control system, which together achieve precise positioning and stable clamping of the work rolls.

[0074] The sensing system is responsible for collecting various status information of the work roll and the positioning and clamping device. The system includes an array of temperature sensors, a strain sensor array, and a displacement sensor array arranged along the axial direction of the work roll. These sensors collect real-time temperature, stress, and displacement field data of the work roll during the laser cladding process, providing a data foundation for multi-physics coupling prediction. The system also includes vibration sensors installed at key locations in the positioning and clamping system. These sensors are positioned in areas prone to vibration, such as the chuck, bearing housing, and V-bracket, to collect vibration signals in real-time, providing a basis for vibration analysis and suppression. The system also includes laser profile scanners and industrial cameras installed at both ends of the work roll to detect deviations in the work roll axis, providing measurement information for alignment and calibration. All these sensors are connected to the control system via a data acquisition card or communication interface to achieve real-time data transmission.

[0075] The temperature sensor array specifically includes two types: infrared temperature sensors and thermal imaging cameras. The infrared temperature sensors are divided into multiple measurement sections along the axial direction of the work roll, preferably 8 to 12 sections. Several infrared temperature sensors are evenly distributed circumferentially within each measurement section, preferably 4 to 6, enabling the measurement of temperature at different circumferential locations within that section. The infrared temperature sensors employ a non-contact temperature measurement method, measuring temperature by detecting infrared radiation emitted from the object's surface, offering advantages such as fast response and a wide measurement range. The thermal imaging camera acquires a two-dimensional temperature distribution image of the work roll surface, providing more comprehensive temperature field information, and is particularly suitable for monitoring the temperature distribution in the molten pool region. The strain sensor array includes two types: strain gauge sensors and fiber Bragg grating strain sensors. The strain gauge sensors operate based on the resistive strain effect; when the material experiences strain, the resistance value of the strain gauge attached to the surface changes, and the strain magnitude can be calculated by measuring the resistance change. The strain gauge sensors and infrared temperature sensors are staggered to avoid mutual interference and to more comprehensively reflect the stress state of the cross-section. Fiber Bragg grating strain sensors measure strain by utilizing the wavelength shift of fiber Bragg gratings. They offer advantages such as strong resistance to electromagnetic interference and the ability to be embedded within structures, enabling the measurement of stress states within the work roll. The displacement sensor array includes laser displacement sensors, with 2 to 4 laser displacement sensors preferably arranged opposite each other at each measurement section. This allows for the measurement of displacement changes on the work roll surface relative to a reference position. The laser displacement sensors employ laser triangulation or laser interferometry principles, offering high precision and high resolution. Vibration sensors preferably utilize triaxial accelerometers, capable of simultaneously measuring vibration acceleration in three orthogonal directions, comprehensively reflecting the spatial characteristics of vibration. These sensors are installed at the chuck position to monitor end vibration, the bearing housing position to monitor bearing system vibration, and the V-bracket position to monitor support system vibration. Multi-point measurements allow for accurate identification of the vibration propagation path and primary sources.

[0076] The execution system is responsible for performing various adjustment actions according to control commands. The execution system includes a three-layer cascaded compensation mechanism for thermal deformation compensation, an active-passive hybrid vibration damping mechanism for vibration suppression, and a multi-field joint alignment mechanism for alignment calibration. These mechanisms work under the unified coordination of the control system to achieve comprehensive control of the working roll's state.

[0077] The three-layer cascaded compensation mechanism is designed to balance the compensation range and accuracy. The mechanism includes a hydraulically driven V-shaped bracket as the coarse compensation layer, a piezoelectric ceramic actuator mounted on the V-shaped bracket as the fine compensation layer, and a Z-axis adjustment mechanism connected to the laser head as the ultra-precise compensation layer. The hydraulically driven V-shaped bracket is connected to a fixed column via a hydraulic cylinder. The hydraulic cylinder extends and retracts under the control of a hydraulic control valve, pushing the V-shaped bracket radially. The hydraulic system features high thrust and long stroke, enabling millimeter-level adjustments, but its response speed is relatively slow. The piezoelectric ceramic actuator utilizes the inverse piezoelectric effect of piezoelectric materials; when voltage is applied, the piezoelectric ceramic undergoes minute expansion and contraction. One end of the piezoelectric ceramic actuator is fixedly connected to the V-shaped bracket, and the other end contacts the surface of the work roller, directly pushing the work roller forward through its own expansion and contraction. Although the stroke of the piezoelectric ceramic actuator is small, it has extremely high resolution and extremely fast response speed, enabling micron-level precision adjustments. The Z-axis adjustment mechanism is connected to the laser head mounting bracket via a guide rail, ensuring the straightness of the movement. The drive end of the Z-axis adjustment mechanism is connected to the output shaft of the voice coil motor. The voice coil motor is a direct-drive linear motor with advantages such as fast response speed, high positioning accuracy, and no mechanical friction, enabling sub-micron level ultra-precision adjustment. The control commands for the three-layer compensation mechanism are output from the predictive compensation control module of the control system. Based on the predicted thermal deformation, the commands are automatically distributed to different levels: large deformations are handled by the hydraulic system, medium deformations by the piezoelectric ceramic, and small deformations by the Z-axis adjustment. This layered strategy ensures both sufficient adjustment range and extremely high final accuracy.

[0078] The active-passive hybrid vibration damping mechanism combines the advantages of both passive and active vibration damping methods. The passive damping component includes a viscoelastic damper and an air spring. The viscoelastic damper is installed between the bearing housing and the foundation, utilizing the internal friction of the viscoelastic material to dissipate vibration energy. It has a certain suppression effect on vibrations of various frequencies, and its damping performance is stable and reliable, requiring no external energy or control. However, its damping characteristics are fixed and cannot be adjusted. The air spring is installed between the V-shaped bracket support base and the foundation, serving as an elastic support element for the vibration isolation platform. The air spring utilizes the compressibility of air to provide elastic force; its stiffness can be changed by adjusting the air pressure. The air spring has a very low natural frequency, providing good isolation for high-frequency vibrations. The active vibration damping component includes a magnetorheological damper and a piezoelectric ceramic actuator array. The magnetorheological damper is installed between the bearing housing and the foundation, arranged in parallel with the viscoelastic damper. By changing the magnetic field strength, the damping coefficient can be adjusted in real time to adapt to different vibration states. The current input terminal of the magnetorheological damper is connected to a PWM controller, which controls the current magnitude through pulse width modulation, thereby controlling the magnetic field strength. An array of piezoelectric ceramic actuators is mounted on a V-shaped bracket support base. When vibration is detected, the piezoelectric ceramic actuators, driven by a control voltage, generate an action force opposite to the direction of vibration, actively canceling the vibration. The drive voltage input terminal of the piezoelectric ceramic actuator is connected to a piezoelectric driver, which provides a high-voltage output to drive the piezoelectric ceramics. Both the PWM controller and the piezoelectric driver receive control signals from the vibration analysis and suppression control module of the control system. Based on the identified vibration type and characteristics, they adjust the damping coefficient and the magnitude of the action force in real time to achieve adaptive vibration control. Passive vibration damping provides basic vibration suppression capabilities, while active vibration damping precisely suppresses specific vibration modes. The combination of the two achieves effective control of broadband vibration.

[0079] The multi-field joint centering mechanism is an innovative design of this invention, integrating three adjustment methods: thermal field, electromagnetic field, and mechanical field. An annular electric heater is embedded in the inner wall of the chucks at both ends of the work roll, making close contact with the work roll surface for efficient heat transfer. The annular electric heater is divided into multiple independently controlled heating zones, preferably eight zones, evenly distributed along the circumference. Each zone is powered by an independent heating controller, allowing independent control of heating power and heating time. By selectively heating certain zones, a temperature gradient can be generated at specific locations on the work roll, using the thermal expansion effect to drive the work roll axis movement. An electromagnet array is mounted on supports on both sides of the V-shaped bracket. The support positions are precisely designed so that the magnetic pole faces of the electromagnets are directly opposite the work roll surface, maintaining an appropriate air gap distance. The electromagnet array preferably includes six electromagnets on each side, distributed axially, each powered by an independent electromagnetic driver. The electromagnetic driver can precisely control the magnitude and direction of the coil current, thereby controlling the electromagnetic attraction. By coordinating the control of the current of each electromagnet, different attraction forces can be generated at different positions on the work roll, achieving non-contact adjustment of the axial position. The servo-driven electric ball screw provides the final precise positioning for mechanical alignment. The nut end of the ball screw is fixedly connected to the movable column, and the screw end is connected to the output shaft of the servo motor via a coupling. The movable column is connected to the fixed base via a linear guide, ensuring the straightness of the movement and repeatability of the positioning. The servo motor receives control signals from the servo motor driver, which performs precise closed-loop position control based on position information fed back from the encoder. The heating controller, electromagnetic driver, and servo motor driver all receive control signals from the alignment optimization control module of the control system. This module calculates the optimal thermal, electromagnetic, and mechanical field parameters based on the detected eccentricity and eccentricity direction using a sequential quadratic programming algorithm, achieving coordinated control of the three fields. This multi-field joint design makes the alignment process both fast and precise, while minimizing the adverse effects on the work rollers.

[0080] The control system acquires sensor data, executes various algorithms, and generates control commands. It includes a data acquisition and processing module, a predictive compensation control module, a vibration analysis and suppression control module, and a centering optimization control module. These modules coordinate with each other to achieve the system's intelligent control function.

[0081] The hardware architecture of the control system includes an industrial PC, motion control card, data acquisition card, and fieldbus. The industrial PC is the core computing unit of the control system, running a real-time operating system to ensure real-time control. The real-time operating system ensures that critical tasks are completed within a specified time, avoiding performance degradation due to operating system scheduling delays. The industrial PC connects to the motion control card and data acquisition card via expansion interfaces to communicate with external devices. The motion control card is dedicated motion control hardware with multi-axis linkage control capabilities, capable of simultaneously controlling multiple servo motors, stepper motors, and other actuators to achieve complex motion trajectories. The motion control card connects to hydraulic control valve drivers, piezoelectric drivers, voice coil motor drivers, PWM controllers, heating controllers, electromagnetic drivers, and servo motor drivers, sending control commands to these drivers. The data acquisition card is responsible for acquiring analog and digital signals, connecting to temperature sensors, strain sensors, displacement sensors, and vibration sensors, converting the sensor output signals into digital signals for processing by the industrial PC. The data acquisition card has multi-channel synchronous acquisition capabilities, with a sampling rate reaching tens of kilohertz, enabling accurate capture of rapidly changing physical quantities. The laser profile scanner and industrial camera communicate with the industrial control computer via a fieldbus. The fieldbus preferably uses the EtherCAT protocol, which is a high-performance industrial Ethernet protocol with advantages such as fast communication speed, good real-time performance, and flexible topology, and can meet the real-time data exchange needs between multiple sensors and actuators.

[0082] The data acquisition and processing module is the foundational layer of the control system, responsible for acquiring raw data from various sensors and performing preprocessing. This module receives analog or digital signals from temperature, strain, displacement, and vibration sensors via a data acquisition card, performs preprocessing operations such as filtering, calibration, and coordinate transformation to remove measurement noise, convert the data into physical quantities, and then transmits the processed data to the upper-level control module. Preprocessing is crucial for improving control accuracy and stability because raw sensor signals often contain various noises and interferences, which can cause controller output jitter if used directly.

[0083] The predictive compensation control module implements thermal deformation prediction and compensation based on multiphysics coupling. This module comprises three sub-units: a partial least squares regression unit, a Kalman filter unit, and a compensation control unit. The partial least squares regression unit is responsible for establishing a regression model between input and output variables, extracting latent variables, and handling multicollinearity issues. The Kalman filter unit is responsible for establishing a state-space model, performing state estimation and prediction based on the system dynamic equations and measurement data, exhibiting optimal estimation characteristics and good noise immunity. The compensation control unit is responsible for calculating compensation control commands based on the predicted thermal deformation, implementing adaptive gain adjustment, and distributing the control commands to the three-layer compensation mechanism. These three units work together to achieve a complete closed loop from data acquisition and model prediction to control output, enabling the system to actively predict and compensate for thermal deformation, significantly improving positioning accuracy and control performance.

[0084] The vibration analysis and suppression control module implements intelligent vibration control based on joint frequency and time domain analysis. This module comprises three sub-units: a feature extraction unit, a support vector machine (SVM) classification unit, and a vibration control unit. The feature extraction unit performs time-domain, frequency-domain, and time-frequency-domain analysis on the acquired vibration signals, calculating time-domain features such as root mean square (RMS) and peak factor; extracting frequency-domain features such as dominant frequency and spectral entropy using Fast Fourier Transform (FFT); and extracting time-frequency-domain features such as energy at each layer using wavelet packet decomposition, constructing a comprehensive feature vector. The SVM classification unit identifies the vibration type based on the feature vector, classifying vibrations into categories such as unbalanced vibration, bearing fault vibration, laser shock vibration, or resonant vibration. The vibration control unit selects the appropriate suppression strategy based on the identified vibration type and generates targeted control commands to send to the active-passive hybrid vibration damping mechanism. The collaborative work of these three units enables the system to intelligently identify and suppress various complex vibrations, effectively improving the stability of the laser cladding process.

[0085] The centering optimization control module implements intelligent centering functionality based on multi-field collaborative optimization. This module comprises three sub-units: a deviation detection unit, a sequential quadratic programming optimization unit, and a multi-field control unit. The deviation detection unit integrates results from multiple detection methods, including laser scanning, machine vision, and strain field analysis, to calculate the eccentricity and eccentricity angle of the work roller axis. The sequential quadratic programming optimization unit establishes a multi-field collaborative optimization model, solving for the optimal thermal, electromagnetic, and mechanical field control parameters under constraints such as accuracy, safety, and energy consumption. The multi-field control unit converts the optimization results into specific control commands, which are sent to the heating controller of the ring electric heater, the electromagnetic actuator of the electromagnet array, and the servo motor driver of the servo electric ball screw, coordinating the effects of the three physical fields. The collaborative work of these three units ensures a fast and accurate centering process while optimizing energy consumption and the impact on the workpiece.

[0086] Reference Figure 10 This invention provides a positioning and clamping method for repairing the surface of large work rolls in steel plants based on laser cladding. The method includes steps such as establishing a multi-physics coupled prediction model, performing multi-step look-ahead prediction, implementing adaptive compensation control, extracting vibration features, classifying and suppressing vibration sources, detecting axis deviation, performing multi-field joint alignment, and controlling the laser cladding process. The specific implementation process is as follows:

[0087] In step S1, a multiphysics coupling prediction model needs to be established first. During laser cladding, large work rolls are subjected to localized heating by the laser beam, resulting in an uneven temperature distribution across the roll body. This uneven temperature distribution induces thermal stress, leading to thermal deformation of the work roll. To accurately predict this thermal deformation, multiple measurement sections need to be arranged axially on the surface of the work roll. Each measurement section is equipped with a temperature sensor, strain sensor, and displacement sensor, which collect real-time data on the temperature, stress, and displacement fields of the work roll. By establishing a three-field coupling prediction model between the temperature, stress, and displacement fields, the physical state changes of the work roll during the laser cladding process can be accurately described.

[0088] In step S11, the sensor array is configured as follows: The sensor array is uniformly arranged axially on the surface of the work roll. One measurement section, The optimal value is 8 to 12 to ensure comprehensive monitoring of the condition across the entire length of the work roll. Several infrared temperature sensors are configured on each measurement section, uniformly distributed along the circumference of the work roll, preferably 4 to 6, to collect temperature field data at that section. The temperature field data reflects the heating effect of the laser beam on the work roll surface and the conduction and diffusion of heat within the roll body. Simultaneously, several strain gauge sensors are configured on each measurement section, staggered from the temperature sensors, preferably 4 to 6, to collect stress field data. The stress field data reflects the distribution of internal stress generated by the work roll under the combined effects of thermal expansion and external constraints. Furthermore, several laser displacement sensors are configured on each measurement section, preferably 2 to 4, arranged opposite each other, to collect displacement field data. The displacement field data reflects the deformation displacement of the work roll under thermal stress.

[0089] In step S12, a three-field coupling relationship is established based on the collected temperature field data, stress field data, and displacement field data. The distribution of the temperature field is mainly determined by three factors: laser power, scanning speed, and substrate temperature. The distribution of the stress field is jointly determined by the temperature field, the coefficient of thermal expansion, and the elastic modulus. When the local temperature of the working roll increases, the material will undergo thermal expansion. However, due to the overall constraint of the working roll, the thermal expansion is restricted, thereby generating thermal stress inside the roll body. The magnitude of the thermal stress is closely related to the temperature change, the coefficient of thermal expansion of the material, and the elastic modulus. The distribution of the displacement field is determined by the stress field, the moment of inertia of the working roll, and its length. Under the action of thermal stress, the working roll will undergo bending deformation. The magnitude of the deformation is directly proportional to the magnitude of the stress, inversely proportional to the moment of inertia of the working roll, and also related to the length of the working roll. By establishing this three-field coupling relationship, the complete physical process from heat input to final deformation can be accurately described, providing a theoretical basis for subsequent prediction and compensation.

[0090] In step S2, multi-step look-ahead prediction is required. Traditional control methods are often passively responsive, compensating only after deformation is detected, which introduces a lag. The multi-step look-ahead prediction method used in this invention can predict the thermal deformation trend over a future period based on current and historical data, thus achieving proactive compensation. Specifically, a combined prediction model is established using partial least squares regression (PLR) combined with a Kalman filter. PLR is a multivariate statistical analysis method, particularly suitable for handling multicollinearity among independent variables. In this invention, there is a strong correlation between temperature, stress, and displacement data at each measurement section; directly using traditional regression methods would lead to model instability. PLR eliminates the influence of multicollinearity by extracting latent variables, enabling the establishment of a more robust prediction model. The Kalman filter, on the other hand, is a recursive algorithm that can optimally estimate the system state based on the system's dynamic model and measurement data, and has good noise immunity. Combining these two methods fully leverages their respective advantages to achieve accurate prediction of future thermal deformation.

[0091] In step S21, the latent variables need to be extracted first. Temperature, stress, and displacement data collected at each measurement section over a historical time period are arranged into an input matrix based on time series, and the corresponding thermal deformation at each moment is arranged into an output matrix. These matrices are then decomposed using a partial least squares regression algorithm to extract the latent variables. One latent variable, The optimal value for is between 3 and 5. These latent variables are linear combinations of the original variables, capable of summarizing the main information in the original data with fewer dimensions, while eliminating the effects of multicollinearity among the original variables. The process of extracting latent variables is essentially a dimensionality reduction process for high-dimensional data, which simplifies the complexity of the model while retaining the main information, thus improving the accuracy of prediction and computational efficiency.

[0092] In step S22, a state-space model is established. The Kalman filter uses a state-space description method to represent the dynamic characteristics of the system. The state equation describes the evolution of the system state over time, and its form is:

[0093] ;

[0094] In this equation, express The state vector at any given time contains the temperature, stress, and displacement values ​​of each measurement section. The dimension of the state vector is equal to the number of measurement sections multiplied by 3. express The state vector at any given time; It is the state transition matrix, which describes the system's transition from state to state. Time's up The natural evolution of time, the elements of the state transition matrix reflect the mutual influence between the state variables; yes The input vector at any given time includes three control inputs: laser power, scanning speed, and rotational angular velocity. These inputs directly affect the thermal state of the work roll. It is the input matrix, which describes the degree of influence of the input quantity on the state change; Process noise represents the uncertainty of the model itself and the influence of external unmodeled disturbances. Process noise is typically assumed to be Gaussian white noise with zero mean. The observation equation describes the relationship between the system state and the measured values, and its form is: ;

[0095] yes The observation vector at any given time contains the actual measurements from the sensor; It is the observation matrix, which describes how state variables are mapped to observations. The form of the observation matrix depends on the arrangement of the sensors and the measurement principle. This refers to observation noise, which represents the uncertainty in sensor measurements. Observation noise is also assumed to be Gaussian white noise with zero mean. By establishing such a state-space model, the Kalman filter algorithm can be used to recursively estimate and predict the system state.

[0096] In step S23, multi-step prediction is performed. Based on the current time... state vector and input vector The future is calculated recursively using the state equation. Step by step The amount of thermal deformation at time t. The specific prediction formula is:

[0097] ;

[0098] in, Indicates prediction Thermal deformation at any given moment, in mm; It refers to the number of prediction steps, that is, how many time steps forward to predict. The preferred value is one that corresponds to a future time of 5 to 10 seconds; This represents the summation operator; Represents the state transition matrix of The exponent indicates the result after... The state transition effect after a certain time step. The physical meaning of this prediction formula is to predict the future... The thermal deformation of the work roll is decomposed into two contributing factors: the first is the natural evolution of the current state in the future, and the second is the cumulative effect of the input at each future moment. This multi-step forward prediction allows for early knowledge of the impending thermal deformation trend of the work roll, providing valuable time for compensation control and significantly improving the lag problem of traditional passive control methods.

[0099] In step S3, adaptive compensation control needs to be implemented. Based on the predicted future thermal deformation, the corresponding compensation control value is calculated, and the actual compensation action is executed through a three-layer cascaded compensation mechanism. The three-layer compensation mechanism includes coarse compensation for the hydraulically driven V-shaped bracket, fine compensation for the piezoelectric ceramic actuator, and ultra-precise compensation for the laser head Z-axis adjustment mechanism. This multi-level compensation strategy can take into account both the requirements of compensation range and compensation accuracy, achieving full-scale compensation from millimeter to submicron levels.

[0100] In step S31, the compensation amount is first calculated. Based on the predicted thermal deformation, the radial compensation displacement is calculated using the compensation control law. The compensation control law adopts a proportional-derivative control form, and its specific expression is as follows:

[0101] ;

[0102] in, express The radial compensation displacement that needs to be performed at any given time, in mm; It is the proportional gain, which represents the speed at which the compensation amount follows the predicted deformation. The larger the proportional gain, the faster the compensation response, but it may cause oscillations. It is a prediction Thermal deformation at any given moment, in mm; It is the differential gain, which represents the degree of response of the compensation amount to the rate of change of deformation. The introduction of differential gain can improve the damping characteristics of the system and suppress oscillations. It is a prediction The thermal deformation at any given time is expressed in mm, and the difference within square brackets represents the rate of change of the predicted deformation. This proportional-derivative control law considers not only the current deformation magnitude but also the deformation trend, enabling a smoother and faster compensation response.

[0103] In step S32, the control gain needs to be adjusted. Fixed proportional and differential gains are difficult to adapt to the complex and variable conditions during laser cladding. This invention uses fuzzy rules to adaptively adjust the control gain. Specifically, based on the predicted magnitude and rate of change of thermal deformation, the proportional gain is dynamically adjusted using pre-set fuzzy rules. and differential gain The numerical values ​​are as follows: When the predicted deformation is large and the rate of change is large, it indicates that thermal deformation is developing rapidly. In this case, the proportional gain should be increased to accelerate the compensation speed. When the predicted deformation is small and the rate of change is small, it indicates that thermal deformation is stabilizing. In this case, the proportional gain should be decreased to avoid overcompensation. The use of fuzzy rules enables the controller to autonomously adjust control parameters according to the actual situation, enhancing the system's adaptability to different operating conditions. This adaptive mechanism effectively improves the robustness of compensation control.

[0104] In step S33, three-layer compensation is performed. The calculated compensation amount is then... The compensation mechanisms are rationally allocated according to their characteristics. The hydraulic system drives the V-shaped bracket to perform millimeter-level coarse compensation. The coarse compensation layer has a large adjustment range, preferably ±10mm, but its response speed is relatively slow, with a response time of approximately 1 to 3 seconds. Coarse compensation is mainly used to eliminate large-scale thermal deformation. The piezoelectric ceramic actuator performs micrometer-level fine compensation. The fine compensation layer has a smaller adjustment range, preferably 100μm, but its response speed is very fast, with a response time of less than 0.1 seconds. Fine compensation is used to eliminate moderate-scale residual deformation. The laser head Z-axis adjustment mechanism performs sub-micrometer-level ultra-precision compensation. The ultra-precision compensation layer has the smallest adjustment range, preferably ±5mm, but it has extremely high resolution, reaching 0.5μm, and an extremely fast response speed, with a response time of less than 0.05 seconds. Ultra-precision compensation is used to eliminate minute residual errors, achieving final high-precision positioning. The cascaded configuration of the three-layer compensation mechanism fully leverages the advantages of each layer, ensuring both sufficient compensation range and extremely high compensation accuracy. This multi-layer design is one of the key technical means for achieving high-performance thermal deformation compensation in this invention.

[0105] In step S4, vibration feature extraction needs to be performed. The work roller generates various forms of vibration during rotation, which can affect the stability and quality of laser cladding. To effectively suppress vibration, it is first necessary to accurately identify the characteristics of the vibration. Vibration sensors are installed at key locations in the positioning and clamping system, including the chuck, bearing housing, and V-bracket, to collect vibration signals. Then, multi-domain analysis is performed on the collected vibration signals to extract time-domain, frequency-domain, and time-frequency-domain features, providing a basis for subsequent vibration source identification and suppression strategy selection.

[0106] In step S41, vibration signals are acquired. An accelerometer is installed at the chuck position to monitor the vibration at the end of the work roll. An accelerometer is installed at the bearing housing position to monitor the vibration state of the bearing system. An accelerometer is installed at the V-bracket position to monitor the vibration response of the support system. These accelerometers are preferably triaxial accelerometers, capable of simultaneously measuring vibration acceleration in three orthogonal directions. Let... express The vibration acceleration value at time t, in m / s² 2 , The time is represented in seconds (s). By continuously acquiring vibration acceleration signals, a complete waveform showing the change of vibration over time can be obtained.

[0107] In step S42, time-domain features are extracted. Time-domain features can directly reflect the statistical characteristics and waveform features of the vibration signal. First, the root mean square value is calculated. The calculation formula is as follows: ;

[0108] in, This represents the root mean square value, in m / s. 2 It reflects the energy level of the vibration signal; This is the number of sampling points, representing the number of data points collected within an analysis window; Indicates the first Vibration acceleration values ​​at each sampling point, in m / s² 2 ; This represents the summation operator, which sums the squared values ​​of the vibration acceleration at all sampling points. This represents the square root operator. A larger root mean square value indicates more severe oscillations. Then, the cradle factor is calculated. The calculation formula is as follows: ;

[0109] in, The peak factor is a dimensionless parameter that reflects the waveform characteristics of the vibration signal. This represents the maximum value of the vibration signal, measured in m / s. 2The magnitude of the peak factor can distinguish different types of vibration; periodic vibrations have relatively small peak factors, while impact vibrations have larger peak factors. By extracting these time-domain features, the nature and severity of the vibration can be preliminarily determined.

[0110] In step S43, frequency domain features are extracted. Frequency domain features reveal the frequency components of the vibration signal, which is crucial for identifying vibration sources, as different vibration sources often possess specific frequency characteristics. For the vibration signal... The spectrum is obtained by performing a fast Fourier transform. In this transformation, Represents the frequency domain spectrum, which is frequency. The function; Frequency is expressed in Hz. The Fast Fourier Transform (FFT) is an efficient algorithm that converts a time-domain signal into a frequency-domain representation, revealing the various frequency components and their amplitudes within the signal. Spectral analysis can identify the dominant frequency. The dominant frequency is defined as the frequency corresponding to the maximum amplitude of the spectral spectrum, measured in Hz. The dominant frequency reflects the main vibration mode and is an important basis for determining the type of vibration. For example, the dominant frequency of vibration caused by work roll imbalance is usually equal to or close to the rotational frequency of the work roll. The spectral entropy also needs to be calculated. The calculation formula is as follows: ;

[0111] in, Spectral entropy is a dimensionless parameter that reflects the degree of dispersion of spectral energy. Represents frequency The normalized energy probability at a given point is calculated by taking the frequency... The square of the spectral amplitude at a given frequency is divided by the sum of the squares of the amplitudes at all frequencies. The logarithmic operator is typically represented by the natural logarithm. A larger spectral entropy indicates a more dispersed distribution of vibrational energy, while a smaller spectral entropy indicates that the vibrational energy is concentrated at certain specific frequencies. The magnitude of spectral entropy can be used to distinguish between single-frequency vibrations and broadband vibrations.

[0112] In step S44, time-frequency domain features are extracted. Time-domain and frequency-domain features can only reflect a single characteristic of a signal in time or frequency, respectively, while time-frequency domain features can simultaneously reflect the variation patterns of the signal in both time and frequency dimensions. This is particularly useful for analyzing non-stationary vibration signals. This invention uses wavelet packet decomposition to extract time-frequency domain features. For vibration signals... conduct Layer wavelet packet decomposition, This indicates the number of wavelet packet decomposition levels, preferably 3 to 5. Wavelet packet decomposition is a multi-resolution analysis method that decomposes a signal into different frequency bands layer by layer, doubling the frequency resolution of the signal after each layer of decomposition. The decomposition process then calculates the... Layer energy The calculation formula is as follows: ;

[0113] in, Indicates the first The energy characteristics of the layer, in m 2 / s 4 This reflects the magnitude of the vibrational energy within that frequency band; Indicates the first Layer Wavelet coefficients, which are the result of wavelet transform, reflect the characteristics of the signal at specific times and frequencies. By calculating the energy distribution of each layer, the energy distribution of the vibration signal in different frequency bands can be obtained, providing a powerful tool for identifying complex vibrations.

[0114] In step S45, a feature vector is constructed. The previously extracted time-domain features (including root mean square value and peak factor), frequency-domain features (including dominant frequency and spectral entropy), and time-frequency-domain features (including wavelet packet energy at each layer) are combined to form the feature vector. The eigenvector has the following form: Among them, RMS is the root mean square value, which reflects the energy of the vibration signal. The larger the value, the more intense the vibration. CF is the peak factor, which reflects the waveform characteristics of the vibration signal. It is dimensionless and distinguishes different types of vibration. The peak factor of periodic vibration is small, while the peak factor of impact vibration is large. The dominant frequency, corresponding to the frequency with the largest spectral amplitude, reflects the main frequency components of the vibration. Different vibration sources have specific frequency characteristics. This feature vector comprehensively reflects multiple characteristics of the vibration signal, providing sufficient information for subsequent vibration source classification and identification. By comprehensively utilizing multi-domain features, the accuracy and reliability of vibration source identification can be significantly improved.

[0115] In step S5, vibration source classification and suppression need to be performed. Different types of vibration have different generation mechanisms and frequency characteristics, requiring corresponding suppression strategies to achieve the best results. This invention first uses a support vector machine to intelligently classify vibration sources, and then selects an appropriate active or passive suppression method, or a hybrid suppression method combining both, based on the identified vibration type.

[0116] In step S51, vibration sources are classified. The constructed feature vectors are... The input is fed into a pre-trained Support Vector Machine (SVM) classifier. SVM is a supervised learning model that learns the mapping relationship between features of historical sample data and vibration types, enabling accurate classification of new vibration signals. The core idea of ​​SVM is to find the optimal classification hyperplane in a high-dimensional feature space, ensuring that samples of different categories are separated by maximum margin. This invention identifies vibrations into four main types: unbalanced vibration, bearing failure vibration, laser shock vibration, and resonant vibration. Unbalanced vibration is caused by uneven mass distribution of the work roll, characterized by a vibration frequency that is an integer multiple of the work roll's rotational frequency. Bearing failure vibration is caused by defects or wear of internal bearing components; its characteristic frequencies are related to the bearing's geometry and rotational speed, including characteristic frequencies of outer ring defects, inner ring defects, and rolling element defects. Laser shock vibration is caused by the instantaneous impact force generated when a laser beam acts on the surface of the work roll; it is characterized by a pulsed waveform containing many high-frequency components. Resonant vibration occurs when the external excitation frequency approaches the system's natural frequency; it is characterized by a significantly increased vibration amplitude and a frequency approaching a certain natural frequency of the system. Accurate identification of vibration types can provide a basis for selecting targeted suppression strategies.

[0117] In step S52, unbalanced vibration is suppressed. Unbalanced vibration mainly occurs in the low-frequency range, preferably less than 20Hz. For this type of vibration, a passive suppression method is first used, which changes the mass distribution of the work roll by adjusting the position of the counterweight, making the center of mass coincide with the axis of rotation, thereby reducing the unbalanced torque at its source. The adjustment of the counterweight is usually completed during the installation stage, and the position and mass of the counterweight need to be determined based on the dynamic balance test results of the work roll. During the cladding process, an active suppression method can also be used, which compensates for the influence of the unbalanced force by adjusting the speed of the servo motor in real time. The speed adjustment compensation algorithm is as follows: ;

[0118] in, express The angular velocity after compensation at any given moment, in rad / s; This indicates the set angular velocity of the working roller, in rad / s. It is the unbalanced compensation coefficient, which is determined by the system identification method and reflects the suppression effect of angular velocity adjustment on unbalanced vibration; This indicates the detected vibration amplitude, in m / s. 2 ; Represents the sine function; This indicates the rotational angular velocity of the work roll, expressed in rad / s. The phase angle, expressed in rad, determines the timing of the compensating force. The algorithm works by fine-tuning the rotation speed in real time based on the magnitude and direction of the unbalanced force during the work roll's rotation, generating a compensating torque opposite to the unbalanced torque, thereby reducing vibration. This active compensation method offers fast response and can track changes in the work roll's state in real time.

[0119] In step S53, bearing vibration is suppressed. Bearing vibration is mainly concentrated in the mid-frequency range, preferably 20 to 200 Hz. For this type of vibration, a passive suppression method is used to increase the system's damping ratio by adding a viscoelastic damper. The viscoelastic damper dissipates vibration energy using the internal friction mechanism of the viscoelastic material. The damping coefficient of the damper can be adjusted by selecting different damping materials, preferably in the range of 5000 to 50000 N·s / m. An active suppression method uses a magnetorheological damper to adjust the system's damping characteristics in real time. The magnetorheological damper is filled with a magnetorheological fluid, a smart material whose apparent viscosity can change rapidly under the influence of a magnetic field. By controlling the strength of the applied magnetic field, the damping force of the magnetorheological damper can be adjusted in real time. The formula for calculating the damping force is: ;

[0120] in, express Damping force at time t, in N; Indicates that the control current varies. The varying damping coefficient, measured in N·s / m, increases with increasing current. This represents the control current applied to the magnetorheological damper coil, expressed in amperes (A). express The vibration velocity at any given time is expressed in m / s, and is the derivative of the vibration displacement with respect to time. Magnetorheological dampers have a very short response time, typically within tens of milliseconds, enabling them to rapidly adjust the damping force according to real-time changes in the vibration state. This adaptive damping adjustment mechanism significantly improves the suppression effect on mid-frequency vibrations.

[0121] In step S54, laser shock vibration is suppressed. Laser shock vibration mainly occurs in the high-frequency range, preferably above 200Hz. For this type of high-frequency vibration, the passive suppression method uses an air spring as the elastic element of the vibration isolation platform. The air spring has a very low natural frequency, preferably 1 to 2Hz. According to vibration isolation theory, the vibration isolation effect is significant when the excitation frequency is much higher than the natural frequency. When the voltage is doubled, the vibration isolation efficiency can reach over 95%. The active suppression method uses an array of piezoelectric ceramic actuators to generate a reverse control force. Piezoelectric ceramic materials exhibit the inverse piezoelectric effect; when a voltage is applied, they undergo mechanical deformation, which generates a driving force. The control law is:

[0122] ;

[0123] in, express The control force output by the actuator at all times, in N, with a negative sign indicating that the direction of the control force is opposite to the disturbance force; The transfer function of the adaptive filter is represented by . It is the Laplace operator, representing a frequency domain variable; the adaptive filter can automatically adjust its frequency response according to the characteristics of the vibration signal, and achieve targeted suppression of different frequency components; express The disturbance force at any given time, measured in nanoseconds (N), is the excitation source causing vibration. This control law employs a feedforward control approach, measuring the disturbance force and generating an equal and opposite control force to actively cancel the vibration. The piezoelectric ceramic actuator has an extremely fast response speed, effectively suppressing high-frequency vibrations. This combined active and passive approach leverages the general effectiveness of passive vibration isolation for broadband vibrations while utilizing the targeted suppression capability of active control for specific frequencies.

[0124] In step S6, axis deviation detection needs to be performed. The alignment accuracy between the axis of the work roll and the laser processing axis directly affects the cladding quality. During laser cladding, the axis of the work roll may shift due to various factors such as thermal deformation, gravity, and support deformation. This invention employs a multi-modal detection method, combining laser scanning, machine vision, and strain field detection to comprehensively detect the axis deviation of the work roll, improving detection accuracy and reliability through multi-source information fusion.

[0125] In step S61, laser scanning detection is performed. Laser profile scanners are installed at both ends of the work roll. The laser profile scanners utilize the principle of laser triangulation, enabling non-contact measurement of the profile coordinates of an object's surface. During the rotation of the work roll, the laser profile scanners continuously acquire the profile point set at the roll end. Assuming the acquired points... The contour point, the first The coordinates of the contour points are as follows: , and The units are all in mm. The value ranges from 1 to . It refers to the number of data collection points, with the best selection. The rotation is 360 degrees, meaning one point is collected for every degree of rotation, thus obtaining the complete circumferential profile of the work roller end. By performing circle fitting on these profile points using the least squares method, the coordinates of the circle center can be determined. The goal of the fitting is to minimize the sum of squared distances from all contour points to the center of the circle. Least squares is a classic parameter estimation method that determines model parameters by minimizing the sum of squared errors; in this invention, it is used to fit the center position of the roller end circle. The eccentricity measured by the laser scanning method is calculated. The calculation formula is as follows: ;

[0126] in, This represents the eccentricity measured by the laser scanning method, in mm. It indicates the distance between the actual center of the circle and the reference axis. This indicates the ideal coordinates of the reference axis at this cross section, in mm. The reference axis is usually defined as a straight line connecting the theoretical centers of the two ends. This represents the square root operation. It calculates the eccentricity angle measured by laser scanning. The calculation formula is as follows: ;

[0127] in, This represents the eccentricity angle measured by the laser scanning method, in rad. It indicates the angle of the eccentricity direction relative to the reference coordinate system. This represents the arctangent function. Determining the eccentricity angle is crucial for subsequent alignment and calibration, as it's necessary to know in which direction the corrective force should be applied.

[0128] In step S62, machine vision inspection is performed. Machine vision inspection, as a supplement to laser scanning inspection, provides another independent measurement method. An image of the work roll end face is captured using an industrial camera. The optical system and image sensor of the industrial camera are precisely calibrated to provide high-resolution images. Digital image processing is performed on the captured image. First, an edge detection algorithm is used to identify the roll end contour. Commonly used edge detection algorithms include the Canny operator and the Sobel operator. These algorithms can detect locations with drastic grayscale changes in the image, thereby identifying object edges. Then, a circle recognition algorithm is used to identify the center coordinates of the circle. A commonly used circle recognition algorithm is the Hough transform, a mapping method from image space to parameter space that can robustly identify geometric shapes in the image. The center coordinates of the circle can be calculated using these image processing algorithms. Furthermore, the eccentricity measured by computer vision method and eccentricity Its calculation formula is similar to that of the laser scanning method, namely: ;

[0129] In these formulas, This represents the eccentricity measured by machine vision, in mm. This represents the eccentricity angle measured by machine vision, in rad. The advantage of machine vision inspection is that it can simultaneously acquire complete image information of the end face, allowing it to measure not only eccentricity but also detect other features of the end face.

[0130] In step S63, strain field detection is performed. Strain field detection indirectly determines axial deviation from the angle of force application. Strain sensors are installed at the contact point between the V-shaped bracket and the work roller to collect the strain distribution of the bracket. When the work roller axis shifts, the contact state between the work roller and the bracket changes, resulting in uneven strain distribution on the bracket. Analyzing the degree of unevenness in strain distribution can determine whether eccentricity exists. Specifically, the off-center loading index is calculated. The calculation formula is as follows: ;

[0131] in, The eccentricity index is a dimensionless parameter that reflects the degree of non-uniformity in strain distribution. This indicates the maximum strain value at the measurement point; This represents the minimum strain value at the measurement point. A large eccentricity index indicates uneven stress distribution and significant eccentricity. The preferred threshold is an eccentricity index greater than 0.15. Although strain field detection cannot directly provide a numerical value for eccentricity, it can provide information to determine whether eccentricity exists, thus providing supplementary information for multi-source information fusion.

[0132] In step S64, the detection results are fused. Since each measurement method has certain measurement errors and applicable conditions, the measurement results of a single method may not be reliable enough. By weighted fusing the detection results from laser scanning, machine vision, and strain field methods, the advantages of each method can be fully utilized, improving the accuracy of the final detection result. The eccentricity after fusion... The calculation formula is: ;

[0133] in, This indicates the eccentricity after fusion, in mm. Indicates the weight of the laser scanning method; This represents the weight of the machine vision method. The weight value is determined based on the measurement accuracy and reliability of each method, and the sum of the weights equals 1. Laser scanning methods typically have higher measurement accuracy and can be assigned a larger weight, making them the preferred choice. It is between 0.6 and 0.7. Similarly, the eccentricity after fusion... The calculation formula is: ;

[0134] This multi-source information fusion method effectively reduces the random errors of a single measurement method, improves the accuracy and robustness of eccentricity detection, and provides reliable input information for subsequent alignment calibration.

[0135] In step S7, multi-field joint alignment needs to be performed. Traditional alignment methods typically rely solely on mechanical adjustments, which have limited accuracy and are time-consuming. This invention proposes an innovative multi-field joint alignment method that comprehensively utilizes the effects of three physical fields: thermal gradient field, electromagnetic force field, and mechanical force field. Through multi-field collaboration, it achieves rapid and high-precision alignment calibration. Simultaneously, a sequential quadratic programming algorithm is used to globally optimize the control parameters of the three physical fields, minimizing alignment time and energy consumption while meeting alignment accuracy requirements.

[0136] In step S71, thermal gradient alignment is initiated. Thermal gradient alignment is an innovative feature of this invention, utilizing the thermal expansion characteristics of the work roll material to achieve alignment adjustment. Based on the eccentricity direction detected in step S64, the annular electric heaters at both ends of the work roll are heated in sections. The annular electric heaters are divided into several independently controlled heating sections, preferably eight sections, each evenly distributed on the circumference. When eccentricity of the work roll in a certain direction is detected, the heating section on the opposite side of the eccentricity direction is heated, causing the temperature on that side to rise. Because the two ends of the work roll are fixed and constrained, the thermal expansion generated by local heating is constrained, thereby generating thermal stress and thermal torque inside the roll body, driving the work roll axis to move in the opposite direction of eccentricity. The resulting temperature difference... The size of the temperature difference needs to be precisely controlled. If the temperature difference is too small, the centering effect will be insignificant; if the temperature difference is too large, it may cause excessive thermal stress or even plastic deformation. Ideally, the temperature difference should be controlled within the range of 5 to 20°C. (Thermal induced deformation amount) The calculation formula is: ;in, This represents the amount of thermally induced deformation, expressed in mm. It indicates the axial offset caused by the thermal gradient. This represents the coefficient of linear expansion of the work roll material, expressed in units of 1 / ℃. For commonly used steel, the coefficient of linear expansion is approximately... / ℃; This indicates the length of the work roll, in mm. It represents the temperature difference, measured in °C, and is the temperature difference between the heated and unheated areas. The value represents the diameter of the work roll, in mm. This formula, derived from beam bending theory, reflects the quantitative relationship between temperature difference, geometric dimensions, and deformation. The advantage of the thermal gradient alignment method is its gentle action, avoiding impact and damage to the work roll. It is particularly suitable for fine-tuning the position of precision workpieces, and its application in work roll repair is highly innovative.

[0137] In step S72, electromagnetic field alignment is initiated. Electromagnetic field alignment applies adjustment force in a non-contact manner. An array of electromagnets is installed on both sides of the V-shaped bracket, with the magnetic pole faces of the electromagnets towards the surface of the work roller but not in direct contact with it, maintaining a certain air gap distance. An adjustable electromagnetic attraction force can be generated by controlling the current in the electromagnet coils. The formula for calculating the electromagnetic attraction force is: ;

[0138] in, This represents electromagnetic attraction force, measured in N (newtons). This indicates the number of turns in the electromagnet coil; the more turns, the stronger the magnetic field. It represents the current passing through the coil, measured in amperes (A). The magnitude of the current directly controls the magnetic field strength. Represents the permeability of free space, its value is H / m is a physical constant; This indicates the relative permeability of the work roll material, which is typically in the range of several hundred to several thousand for ferromagnetic materials. Represents the area of ​​magnetic poles, in meters (m²). 2 The larger the area of ​​the magnetic poles, the greater the attraction. The air gap distance, expressed in meters (m), represents the distance between the magnetic pole face of the electromagnet and the surface of the work roll. A smaller air gap results in a stronger magnetic field. This formula shows that the electromagnetic attraction is directly proportional to the square of the current and inversely proportional to the square of the air gap, giving the electromagnetic force high controllability and sensitivity. By independently controlling the current of each electromagnet, different magnitudes of attraction can be generated at different positions around the work roll, thus achieving precise adjustment of the work roll's axial position. Electromagnetic field alignment has a very fast response speed, with the electromagnetic force building up in milliseconds, enabling rapid position adjustment—a significant advantage of electromagnetic alignment over traditional mechanical alignment.

[0139] In step S73, mechanical alignment is initiated. Mechanical alignment achieves final precise positioning through direct mechanical displacement. A servo-driven electric ball screw drives the V-shaped bracket for position adjustment. The ball screw converts the rotational motion of the servo motor into the linear motion of the bracket, offering advantages such as high transmission accuracy, good rigidity, and low friction. Precise positioning is achieved through closed-loop position control, the control law of which is: ;

[0140] in, express The control value at any given moment, expressed in mm, is the distance the ball screw needs to move. This represents the proportional gain, which determines the controller's response speed to the current error. express The eccentricity at any given moment, expressed in mm, is the deviation between the target position and the actual position. This represents the integral gain. The integral term is used to eliminate steady-state errors and ensure that the target position can be accurately reached in the end. This represents the summation operator, for values ​​from 1 to... Sum of all historical errors; This represents the differential gain. The differential term is used to predict the trend of error changes and improve the dynamic response speed and stability of the system. express The eccentricity at any given time is expressed in mm. This control law is a classic PID controller, which is widely used in industrial control due to its simplicity and effectiveness. The proportional term ensures the response speed, the integral term ensures the steady-state accuracy, and the derivative term ensures the system stability. The synergistic effect of these three terms enables rapid and precise positioning of the machine.

[0141] In step S74, multi-field synergistic optimization is performed. Using any single centering method has its limitations: thermal gradient centering has a slow but gentle effect, electromagnetic centering has a fast response but limited force magnitude and range, and mechanical centering offers high precision but may cause mechanical impact to the workpiece. Using multiple fields synergistically can fully leverage their respective advantages and achieve complementary benefits. However, how to rationally allocate the influence of the three physical fields is a complex optimization problem. This invention establishes a multi-field synergistic optimization model, with the objective function being: ;

[0142] in, This represents the objective function to be optimized, with the goal of minimizing the function. The weighting coefficient represents the eccentricity, which reflects the importance of centering accuracy. The larger the weight, the more importance is placed on accuracy. This represents the residual eccentricity after centering, in mm, and ideally should be close to zero. The weighting coefficient of the thermal field reflects the importance of heating energy consumption; This indicates the heating power of each heating zone, in watts (W). This represents the sum of the squares of the total heating power; the smaller this term is, the lower the heating energy consumption. The weighting coefficients representing the electromagnetic field; This represents the current in each electromagnet, expressed in amperes (A). This represents the sum of the squares of the total electromagnetic coil current; Represents the weighting coefficients of the mechanical field; This indicates the displacement of each mechanical actuator, in mm. This represents the sum of squares of the total mechanical displacement. By adjusting the weighting coefficients, different balances can be achieved between centering accuracy, energy consumption, and mechanical impact. A sequential quadratic programming algorithm is used to solve this optimization problem. Sequential quadratic programming is an efficient algorithm for solving nonlinear constrained optimization problems. Its basic idea is to transform the nonlinear optimization problem into a series of quadratic programming subproblems for solution, and then iteratively approximate the optimal solution. This algorithm can handle complex constraints, including eccentricity not exceeding allowable values, temperature difference not exceeding safety limits, electromagnetic force not exceeding rated values, and mechanical displacement not exceeding travel limits. The optimal heating power can be obtained through optimization. Electromagnetic current and mechanical displacement This invention achieves multi-field coordinated control. This multi-field coordinated method based on optimization theory is an important innovation of the invention. It can minimize energy consumption and mechanical impact while ensuring centering accuracy, thereby improving the overall performance of the centering process.

[0143] In step S8, laser cladding process control is performed. In the actual laser cladding repair process, the aforementioned multiphysics coupling prediction and compensation, vibration analysis and suppression, and multi-field joint alignment technologies are executed simultaneously in parallel, forming a complete closed-loop control system. Specifically, the multiphysics coupling prediction model established in steps S1 to S3 continuously predicts the thermal deformation trend of the work roll, and a three-layer compensation mechanism performs compensation actions in real time to dynamically offset the impact of thermal deformation on positioning accuracy, ensuring that the laser beam and the work roll surface always maintain the correct relative positional relationship. The vibration analysis model established in steps S4 to S5 identifies vibration types in real time, and a hybrid active-passive vibration damping mechanism performs targeted vibration suppression, effectively reducing the adverse effects of vibration on cladding quality, keeping the molten pool stable, and ensuring uniform cladding layer thickness. The multimodal detection system established in steps S6 to S7 continuously monitors the positional deviation of the work roll axis, and a multi-field joint alignment system dynamically maintains alignment accuracy, ensuring that the coaxiality of the work roll axis and the laser processing axis always meets the requirements. The coordinated operation of these three subsystems constitutes an intelligent adaptive control system that can cope with various complex disturbances during the laser cladding process, ensure the stability of the repair process and the consistency of repair quality, and significantly improve the technical level of laser cladding repair of large work rolls.

[0144] Example 1: This example describes the use of a positioning and clamping system and method based on laser cladding to repair the surface of a large work roll in a steel plant, specifically for the repair of a hot-rolled work roll. The work roll is 5800mm long, 800mm in diameter, and made of high-chromium cast iron. During use, localized wear and spalling defects occurred on its surface, with a wear depth of approximately 2.5mm, requiring surface repair using laser cladding technology.

[0145] In this embodiment, the sensing system is configured as follows: Ten measurement sections are evenly arranged along the axial direction of the work roll, with a spacing of 580 mm between them. Each measurement section is equipped with five infrared temperature sensors, uniformly distributed circumferentially, for collecting temperature field data. The measurement range of the infrared temperature sensors is 0 to 1200℃, and the measurement accuracy is ±2℃. Each measurement section is equipped with five strain gauge sensors, staggered from the temperature sensors, for collecting stress field data. The measurement range of the strain gauge sensors is ±5000με, and the measurement accuracy is ±1με. Each measurement section is equipped with three laser displacement sensors, arranged opposite each other, for collecting displacement field data. The measurement range of the laser displacement sensors is ±15 mm, and the measurement accuracy is ±0.01 mm. Three-dimensional accelerometers are installed at the chuck, bearing housing, and V-bracket positions to collect vibration signals. The measurement range of the accelerometers is 0 to 100 m / s². 2 The frequency response range is 0.5 to 5000 Hz. Laser profile scanners are installed at both ends of the work roll to detect axial deviation; the measurement accuracy of the laser profile scanners is ±0.005 mm.

[0146] The execution system is configured as follows: In the three-layer cascaded compensation mechanism, the coarse compensation stroke of the hydraulically driven V-shaped bracket is ±12mm, the response time is 2.5 seconds, and the control accuracy is ±0.5mm. The fine compensation stroke of the piezoelectric ceramic actuator is ±120μm, the response time is 0.08 seconds, and the control accuracy is ±0.5μm. The ultra-precise compensation stroke of the laser head Z-axis adjustment mechanism is ±6mm, the response time is 0.04 seconds, and the control accuracy is ±0.3μm. In the active-passive hybrid vibration reduction mechanism, the damping coefficient of the viscoelastic damper is 15000N·s / m, the natural frequency of the air spring is 1.5Hz, the adjustable damping coefficient range of the magnetorheological damper is 5000 to 40000N·s / m, the response time is 30ms, and the piezoelectric ceramic actuator array has a total of 12 actuators, with a maximum output force of 500N per actuator and a response time of 20ms. In the multi-field joint centering mechanism, the annular electric heater is divided into 8 independently controlled zones, with a maximum heating power of 3000W per zone. The maximum temperature difference between the heated and unheated zones is controlled within a range of 5 to 18℃. Each side of the electromagnet array is equipped with 6 electromagnets, with a maximum attraction force of 1200N per electromagnet, a control current range of 0 to 8A, and an air gap distance of 2mm. The servo-driven electric ball screw has a stroke of ±25mm, a positioning accuracy of ±0.005mm, and a repeatability of ±0.002mm. The laser cladding process parameters are set as follows: laser power 3500W, scanning speed 8mm / s, powder feeding rate 15g / min, protective gas flow rate 18L / min, and working roller rotational angular velocity 0.15rad / s, approximately 1.43 rpm.

[0147] The specific implementation steps are as follows: Step S1: Establish a multi-physics coupling prediction model. Step S11: Configure a sensor array, uniformly arranging 10 measurement sections along the axial direction of the working roller surface. At each measurement section, configure 5 infrared temperature sensors to collect temperature field data, 5 strain gauge sensors to collect stress field data, and 3 laser displacement sensors to collect displacement field data. Step S12: Establish a coupling prediction model, establishing the three-field coupling relationship based on the collected temperature field data, stress field data, and displacement field data. The temperature field is determined by a laser power of 3500W, a scanning speed of 8mm / s, and a substrate temperature of 25℃. The stress field is determined by the temperature field and the coefficient of thermal expansion of 1.2×10⁻⁻⁻⁶. 5 The displacement field is determined by the temperature and elastic modulus of 210 GPa, and is determined by the stress field and the moment of inertia of the working roller of 2.01 × 10⁻⁶ GPa. 8 mm 4The length is determined to be 5800mm. Step S2: Perform multi-step look-ahead prediction. Step S21: Extract latent variables. The historical data of temperature, stress, and displacement at each measurement section are used to form an input matrix, and the corresponding thermal deformation is used to form an output matrix. Four latent variables are extracted using partial least squares regression. Step S22: Establish a state-space model and construct the state equation and observation equation of the Kalman filter. Step S23: Perform multi-step prediction. Based on the current state vector and input vector, the thermal deformation at time k=8 steps in the future is recursively calculated using the state equation. The prediction time is 8 seconds. Step S3: Implement adaptive compensation control. Step S31: Calculate the compensation amount. Based on the predicted thermal deformation, the radial compensation displacement is calculated using the compensation control law. The initial proportional gain Kp=0.85 and differential gain Kd=0.12. Step S32: Adjust the control gain. Based on the magnitude and rate of change of the predicted thermal deformation, the proportional gain and differential gain are adaptively adjusted using fuzzy rules. When the predicted deformation is greater than 1.5 mm and the rate of change is greater than 0.3 mm / s, the proportional gain is adjusted to 0.92 and the differential gain is adjusted to 0.15; when the predicted deformation is less than 0.5 mm and the rate of change is less than 0.1 mm / s, the proportional gain is adjusted to 0.78 and the differential gain is adjusted to 0.09. Step S33: Perform three-layer compensation, allocating the compensation amount to the three-layer compensation mechanism. The V-shaped bracket is driven by the hydraulic system to perform millimeter-level coarse compensation, the piezoelectric ceramic actuator to perform micrometer-level fine compensation, and the laser head Z-axis adjustment mechanism to perform sub-micrometer-level ultra-fine compensation. Step S4: Perform vibration feature extraction. Step S41: Collect vibration signals. Accelerometers are installed at the chuck, bearing housing, and V-shaped bracket positions to collect vibration acceleration signals at a sampling frequency of 10 kHz. Step S42: Extract time-domain features, calculating the root mean square value and peak factor. Step S43: Extract frequency-domain features. The vibration signal is subjected to a fast Fourier transform to obtain the spectrum, the dominant frequency is identified, and the spectral entropy is calculated. Step S44: Extract time-frequency domain features, perform 4-level wavelet packet decomposition on the vibration signal, and calculate the energy of the 4th level. Step S45: Construct a feature vector, combining time-domain features, frequency-domain features, and time-frequency domain features to form a feature vector for vibration source classification. Step S5: Perform vibration source classification and suppression. Step S51: Classify vibration sources, inputting the feature vector into a support vector machine classifier to identify vibration types. Step S52: For identified unbalanced vibrations, passive suppression is achieved by adjusting the position of the counterweight, and active suppression is achieved by adjusting the speed of the servo motor. The unbalance compensation coefficient is set to 0.08. Step S53: For bearing vibrations, passive suppression is achieved using a viscoelastic damper with a damping coefficient of 15000 N·s / m, and active suppression is achieved by adjusting the damping coefficient using a magnetorheological damper, with a control current range of 0 to 6 A. Step S54: For laser shock vibrations, passive suppression is achieved using an air spring vibration isolation platform with a natural frequency of 1.5 Hz, and active suppression is achieved by generating a reverse force using a piezoelectric ceramic actuator.Step S6: Perform axis deviation detection. Step S61: Perform laser scanning detection. Install laser profile scanners at both ends of the working roll. During the rotation of the working roll, collect the roll end profile point set. The number of collected points n=360. Fit the center coordinates of the circle using the least squares method to calculate the eccentricity and eccentricity angle. Step S62: Perform machine vision detection. Capture the end face image of the roll using an industrial camera. Use edge detection and circle recognition algorithms to identify the center coordinates of the circle and calculate the eccentricity and eccentricity angle. Step S63: Perform strain field detection. Collect the strain distribution at the V-shaped bracket contact point and calculate the off-center loading index to determine if eccentricity exists. Step S64: Fuse the detection results. Weight the detection results from the laser scanning, machine vision, and strain field methods and fuse them. The weight for the laser scanning method is w1=0.65, and the weight for the machine vision method is w2=0.35. Step S7: Perform multi-field joint alignment. Step S71: Start thermal gradient centering, and heat the annular electric heaters at both ends of the working roll in sections according to the eccentric direction. The temperature difference is set to 12℃ and the linear expansion coefficient is 1.2×10⁻. 5 / ℃. Step S72: Start electromagnetic field alignment. Electromagnetic attraction is generated by the electromagnet array on both sides of the V-shaped bracket. The number of coil turns N=800, and the control current I=5A. Step S73: Start mechanical field alignment. The V-shaped bracket is adjusted in position by driving it with a servo electric ball screw. Position closed-loop control is adopted, with proportional gain Kp=1.2, integral gain Ki=0.15, and differential gain Kd=0.08. Step S74: Perform multi-field collaborative optimization. Establish the optimization objective function and use the sequential quadratic programming algorithm to solve for the optimal heating power, electromagnetic current, and mechanical displacement. The eccentricity weight coefficient λ1=0.6, the thermal field weight coefficient λ2=0.15, the electromagnetic field weight coefficient λ3=0.15, and the mechanical field weight coefficient λ4=0.1. Step S8: Perform laser cladding process control. During the laser cladding process, multi-physics field coupling prediction and compensation, vibration analysis and suppression, and multi-field joint alignment are performed simultaneously to form a complete closed-loop control system to ensure the stability and quality consistency of the repair process.

[0148] Example 2 differs from Example 1 in the sensor configuration and control parameter settings. Eight measurement sections are evenly arranged along the working roller axis, each equipped with four infrared temperature sensors, four strain gauge sensors, and two laser displacement sensors. In the three-layer cascaded compensation mechanism, the coarse compensation stroke of the hydraulically driven V-shaped bracket is ±10mm, the fine compensation stroke of the piezoelectric ceramic actuator is ±100μm, and the ultra-precise compensation stroke of the laser head Z-axis adjustment mechanism is ±5mm. In the active-passive hybrid vibration damping mechanism, the damping coefficient of the viscoelastic damper is 12000 N·s / m, the natural frequency of the air spring is 1.8Hz, and the adjustable damping coefficient range of the magnetorheological damper is 4000 to 35000 N·s / m. The annular electric heater is divided into eight zones, with the maximum temperature difference between the heated and unheated zones controlled within a range of 6 to 15℃. The laser power is 3200W, the scanning speed is 7mm / s, and the working roller rotational angular velocity is 0.12rad / s. Three latent variables are extracted in step S21. In step S23, the number of prediction steps is k=6 when performing multi-step prediction. In step S31, the initial proportional gain Kp=0.80 and the differential gain Kd=0.10. In step S64, the weights for the laser scanning method are w1=0.70 and the machine vision method is w2=0.30. In step S71, the temperature difference is set to 10℃. In step S73, the proportional gain Kp=1.0, the integral gain Ki=0.12, and the differential gain Kd=0.06. In step S74, the eccentricity weighting coefficients are λ1=0.55, the thermal field weighting coefficient λ2=0.20, the electromagnetic field weighting coefficient λ3=0.15, and the mechanical field weighting coefficient λ4=0.10. Other steps are the same as in Example 1.

[0149] Example 3 differs from Example 1 in the sensor configuration and control parameter settings. Twelve measurement sections are evenly arranged along the working roller axis, each equipped with six infrared temperature sensors, six strain gauge sensors, and four laser displacement sensors. In the three-layer cascaded compensation mechanism, the coarse compensation stroke of the hydraulically driven V-shaped bracket is ±15mm, the fine compensation stroke of the piezoelectric ceramic actuator is ±150μm, and the ultra-precise compensation stroke of the laser head Z-axis adjustment mechanism is ±8mm. In the active-passive hybrid vibration damping mechanism, the damping coefficient of the viscoelastic damper is 18000 N·s / m, the natural frequency of the air spring is 1.2Hz, and the adjustable damping coefficient range of the magnetorheological damper is 6000 to 45000 N·s / m. The annular electric heater is divided into eight zones, with the maximum temperature difference between the heated and unheated zones controlled within a range of 8 to 20℃. The laser power is 3800W, the scanning speed is 9mm / s, and the working roller rotational angular velocity is 0.18rad / s. Five latent variables are extracted in step S21. In step S23, the number of prediction steps is k=10 when performing multi-step prediction. In step S31, the initial proportional gain Kp=0.90 and the differential gain Kd=0.14. In step S64, the weights for the laser scanning method are w1=0.60 and the machine vision method is w2=0.40. In step S71, the temperature difference is set to 15℃. In step S73, the proportional gain Kp=1.4, the integral gain Ki=0.18, and the differential gain Kd=0.10. In step S74, the eccentricity weighting coefficient λ1=0.65, the thermal field weighting coefficient λ2=0.12, the electromagnetic field weighting coefficient λ3=0.13, and the mechanical field weighting coefficient λ4=0.10. Other steps are the same as in Example 1.

[0150] Comparative Example 1 uses a traditional positioning and clamping method, employing only a single-layer hydraulic compensation mechanism. It does not use a multi-physics coupling prediction model or perform multi-step look-ahead prediction, but instead uses passive compensation control, meaning compensation is only performed after thermal deformation is detected. The compensation control uses simple proportional control with a fixed proportional gain of 0.5. Other equipment configurations and process parameters are the same as in Example 1.

[0151] Comparative Example 2 differs from Example 1 in that it does not perform the vibration feature extraction and vibration source classification and suppression steps. It only adopts passive vibration reduction measures, that is, it only uses viscoelastic dampers and air springs, and does not adopt active vibration reduction measures such as magnetorheological dampers and piezoelectric ceramic actuators. Other steps and parameters are the same as in Example 1.

[0152] Comparative Example 3 differs from Example 1 in that the centering calibration only uses mechanical field centering, without thermal gradient field centering or electromagnetic field centering, and does not perform multi-field collaborative optimization. Other steps and parameters are the same as in Example 1.

[0153] Comparative Experiment: To verify the implementation effects of the examples and comparative examples, the following comparative experiment was designed to conduct performance comparison tests on examples 1 to 3 and comparative examples 1 to 3.

[0154] Experiment 1: Positioning Accuracy Test; The experimental method is based on GB / T1184-1996 Standard for Shape and Position Tolerances and ISO230-2 General Rules for Machine Tool Inspection, Part 2: Determination of Positioning Accuracy and Repeatability of CNC Axes. During the experiment, a laser interferometer measurement system is used to continuously monitor the relative positional deviation between the laser head and the surface of the work roll during laser cladding. 100 measurement points are used, with a measurement interval of 30 seconds. The positioning accuracy index, i.e., the maximum value of the positional deviation among all measurement points, is calculated. Experiment 2: Thermal Deformation Compensation Accuracy Test; The experimental method is based on ISO230-3 General Rules for Machine Tool Inspection, Part 3: Determination of Thermal Effects. During the experiment, a displacement sensor is used to monitor the thermal deformation of the work roll in real time during laser cladding, while the compensation amount of the compensation system is recorded. The residual thermal deformation amount, i.e., the absolute value of the difference between the actual thermal deformation amount and the compensation amount, is calculated. 120 measurement points are used, with a measurement interval of 25 seconds. The thermal deformation compensation accuracy index, i.e., the root mean square value of the residual thermal deformation amount among all measurement points, is calculated. Experiment 3: Vibration Amplitude Test; The experimental method is based on ISO 10816-3 Evaluation of mechanical vibration of rotating machinery – Part 3 Industrial machinery. During the experiment, acceleration sensors were installed at key locations in the positioning and clamping system, including the chuck, bearing housing, and V-bracket, to continuously monitor vibration acceleration signals during laser cladding. The measurement time was 30 minutes, and the sampling frequency was 10kHz. The vibration amplitude index, i.e., the root mean square value of the vibration acceleration signal, was calculated. Experiment 4: Alignment Accuracy Test; The experimental method is based on GB / T 1184-1996 Measurement Method of Coaxiality Tolerance in the Standard for Shape and Position Tolerances. During the experiment, a laser profile scanner and machine vision system were used to continuously monitor the eccentricity between the work roller axis and the laser processing axis during laser cladding. 150 measurement points were used, with a measurement interval of 20 seconds. The alignment accuracy index, i.e., the maximum value of the eccentricity among all measurement points, was calculated. Experiment 5: Cladding Layer Thickness Uniformity Test; The experimental method follows the JB / T9195-1999 standard for laser cladding technology. During the experiment, after laser cladding, 30 measurement points were uniformly selected axially on the surface of the work roll, and 12 points were uniformly selected circumferentially at each measurement point. An ultrasonic thickness gauge was used to measure the cladding layer thickness, for a total of 360 points. The cladding layer thickness uniformity index was calculated, expressed as the ratio of the standard deviation of thickness to the average thickness; a smaller value indicates better uniformity. Experiment 6: Cladding Layer Surface Roughness Test; The experimental method follows the GB / T1031-2009 standard for surface roughness parameters and their values. During the experiment, after laser cladding, 15 measurement sections were uniformly selected axially on the surface of the work roll, and 8 measurement points were uniformly selected circumferentially at each measurement section. A surface roughness Ra value was measured using a surface roughness meter, for a total of 120 points. The surface roughness index was calculated, i.e., the average Ra value of all measurement points.Experiment 7: Compensation Response Time Test; During the experiment, a step disturbance was applied during laser cladding. A high-speed data acquisition system was used to synchronously record the time of disturbance occurrence, the time when the compensation system detected the deviation, and the time when the compensation was completed. The experiment was repeated 50 times, and the compensation response time, i.e., the average time from the detection of the deviation to the completion of the compensation, was calculated.

[0155] Specific experimental results are as follows: Figures 1-8 As shown. Figure 1 The experimental procedure followed the GB / T 1184-1996 standard for shape and position tolerances. During the laser cladding process, a laser interferometer measurement system was used to continuously monitor the relative positional deviation between the laser head and the surface of the work roll. 100 measurement points were used, with a measurement interval of 30 seconds. The positioning accuracy index was the maximum positional deviation among all measurement points. The centering accuracy test followed the coaxiality tolerance measurement method. A laser profile scanner and machine vision system were used to continuously monitor the eccentricity between the work roll axis and the laser processing axis. 150 measurement points were used, with a measurement interval of 20 seconds. The centering accuracy index was the maximum eccentricity among all measurement points.

[0156] from Figure 1 As can be seen from sub-figure (a), the positioning accuracy of Example 1 is 0.025 mm, Example 2 is 0.032 mm, and Example 3 is 0.021 mm, while the positioning accuracy of Comparative Example 1 is 0.185 mm, Comparative Example 2 is 0.028 mm, and Comparative Example 3 is 0.030 mm. Figure 1 As shown in sub-figure (b), the centering accuracy of Example 1 is 0.015 mm, Example 2 is 0.019 mm, Example 3 is 0.012 mm, Comparative Example 1 is 0.018 mm, Comparative Example 2 is 0.016 mm, and Comparative Example 3 is 0.095 mm. These data indicate that the positioning accuracy of Examples 1-3, which use multi-physics field coupling prediction and a three-layer cascaded compensation mechanism, is improved by more than 86% compared to Comparative Example 1, which only uses single-layer hydraulic compensation. This invention establishes a three-field coupling prediction model of temperature field, stress field, and displacement field, and uses a partial least squares regression algorithm combined with a Kalman filter for multi-step look-ahead prediction. It can predict the trend of thermal deformation in advance and achieve full-scale accurate compensation through millimeter-level coarse compensation of hydraulically driven V-shaped bracket, micrometer-level fine compensation of piezoelectric ceramic actuator, and submicrometer-level ultra-precise compensation of laser head Z-axis adjustment mechanism. Compared to Comparative Example 3 which only uses mechanical field alignment, Examples 1-3 show an improvement in alignment accuracy of over 87%. This is due to the innovative use of a joint alignment mechanism of three physical fields—thermal gradient field, electromagnetic field, and mechanical field—in this invention. The control parameters of the three fields are globally coordinated and optimized through a sequential quadratic programming algorithm, minimizing alignment time and energy consumption while ensuring alignment accuracy.

[0157] Figure 2During the experiment, the thermal deformation compensation accuracy test was conducted according to ISO 230-3 General Rules for Machine Tool Inspection, Part 3: Determination of Thermal Effects. Displacement sensors were used to monitor the thermal deformation of the work roll in real time during laser cladding, and the compensation amount of the compensation system was recorded simultaneously. The residual thermal deformation amount, i.e., the absolute value of the difference between the actual thermal deformation amount and the compensation amount, was calculated. 120 measurement points were used, with a measurement interval of 25 seconds. The thermal deformation compensation accuracy index was the root mean square value of the residual thermal deformation amount at all measurement points. The compensation response time test involved applying a step-type disturbance during laser cladding. A high-speed data acquisition system was used to simultaneously record the time of disturbance occurrence, the time when the compensation system detected the deviation, and the time when the compensation was completed. After repeating the experiment 50 times, the average time from the detection of the deviation to the completion of the compensation was calculated. Figure 2 As shown in sub-figure (a), the thermal deformation compensation accuracy of Example 1 is 0.018 mm, Example 2 is 0.023 mm, and Example 3 is 0.015 mm, while the thermal deformation compensation accuracy of Comparative Example 1 is 0.152 mm, Comparative Example 2 is 0.019 mm, and Comparative Example 3 is 0.020 mm. Figure 2 As shown in subplot (b), the compensation response time for Example 1 is 0.35 seconds, for Example 2 it is 0.42 seconds, and for Example 3 it is 0.31 seconds. The compensation response time for Comparative Example 1 is 2.85 seconds, for Comparative Example 2 it is 0.38 seconds, and for Comparative Example 3 it is 0.36 seconds. These results demonstrate that Examples 1-3, employing multi-step look-ahead prediction and adaptive compensation control, achieve a thermal deformation compensation accuracy of over 88% and a compensation response time of over 88% compared to Comparative Example 1, which uses passive simple proportional control. The Kalman filter state-space model used in this invention can recursively calculate the thermal deformation amount over the next 5 to 10 seconds based on the current state vector and input vector, achieving active predictive compensation rather than traditional passive response compensation. Simultaneously, through fuzzy rules, the proportional gain and differential gain are adaptively adjusted according to the magnitude and rate of change of the predicted thermal deformation, enabling the compensation control law to adapt to the complex and variable conditions during laser cladding. The design of the three-layer cascaded compensation mechanism fully leverages the advantages of the hydraulic system (large thrust and long stroke), piezoelectric ceramics (fast response and high precision), and voice coil motors (frictionless and ultra-precise), achieving full-scale precise control with both sufficient compensation range and extremely high compensation accuracy.

[0158] Figure 3During the experiment, vibration amplitude testing was conducted according to ISO 10816-3, the standard for evaluating mechanical vibration of rotating machinery. Triaxial acceleration sensors were installed at key locations in the positioning and clamping system, including the chuck, bearing housing, and V-bracket. Vibration acceleration signals were continuously monitored during the laser cladding process for 30 minutes at a sampling frequency of 10 kHz. The vibration amplitude index was the root mean square value of the vibration acceleration signal. Surface roughness testing was conducted according to GB / T 1031-2009, the standard for surface roughness parameters and their numerical values. After laser cladding, 15 measurement sections were evenly selected axially on the surface of the work roll. Eight measurement points were evenly selected circumferentially on each measurement section. The surface roughness Ra value was measured using a surface roughness meter. A total of 120 points were measured, and the average Ra value of all measurement points was calculated. Figure 3 As can be seen from subplot (a), the vibration amplitude of Example 1 is 0.82 m / s. 2 Example 2 shows a speed of 0.95 m / s. 2 Example 3 is 0.75 m / s 2 Comparative Example 1 is 0.88 m / s 2 The vibration amplitude of Comparative Example 2 is as high as 2.35 m / s. 2 Comparative Example 3 is 0.85 m / s 2 .from Figure 3 As shown in sub-image (b), the surface roughness of Example 1 is 1.85 μm, Example 2 is 2.12 μm, Example 3 is 1.68 μm, Comparative Example 1 is 4.25 μm, Comparative Example 2 is 2.95 μm, and Comparative Example 3 is 2.38 μm. Compared with Comparative Example 2, which only uses passive vibration reduction, the vibration amplitude of Examples 1-3, which uses vibration feature extraction and classification and active-passive hybrid suppression, is reduced by more than 65%. This invention innovatively uses time-domain, frequency-domain, and time-frequency-domain joint analysis methods to extract vibration features. By calculating time-domain features such as root mean square value and peak factor, and performing fast Fourier transform to extract frequency-domain features such as main frequency and spectral entropy, and performing wavelet packet decomposition to extract time-frequency-domain features such as energy at each layer, a comprehensive feature vector is constructed. Then, a support vector machine classifier is used to accurately identify the vibration as unbalanced vibration, bearing failure vibration, laser shock vibration, or resonance vibration, and corresponding active-passive hybrid suppression strategies are adopted for different types of vibration. The viscoelastic damper and air spring in the passive vibration damping section provide basic vibration suppression capability, while the magnetorheological damper in the active vibration damping section achieves precise suppression of specific vibration modes by adjusting the damping coefficient in real time and generating a reverse control force through a piezoelectric ceramic actuator. Effective vibration control directly improves the stability of the cladding process, keeps the molten pool stable, and reduces surface roughness. The surface roughness improvement in Examples 1-3 compared to Comparative Example 1 reaches over 56%.

[0159] Figure 4During the experiment, the uniformity of the cladding layer thickness was tested according to the JB / T 9195-1999 standard for laser cladding technology. After laser cladding, 30 measurement points were uniformly selected axially on the surface of the work roll, and 12 positions were uniformly selected circumferentially at each measurement point. An ultrasonic thickness gauge was used to measure the cladding layer thickness at a total of 360 points. The uniformity index of the cladding layer thickness was expressed as the ratio of the standard deviation of the thickness to the average thickness; the smaller the value, the better the uniformity. The surface roughness testing method was the same as... Figure 3 Same as above. From Figure 4 As shown in subplot (a), the cladding layer thickness uniformity is 3.2% in Example 1, 3.8% in Example 2, and 2.9% in Example 3, while the thickness uniformity is only 12.5% ​​in Comparative Example 1, 5.8% in Comparative Example 2, and 4.6% in Comparative Example 3. From... Figure 4 Subplot (b) shows the distribution trend of surface roughness for each sample. The surface roughness of Examples 1-3 is significantly lower than that of the comparative example. Experimental results indicate that Examples 1-3, through the comprehensive application of multiphysics coupling prediction compensation, vibration suppression, and multi-field joint alignment, improve the uniformity of cladding layer thickness by more than 74% compared to Comparative Example 1. The multiphysics coupling prediction model of this invention accurately predicts the thermal deformation trend of the work roll and compensates in real time through a three-layer cascaded compensation mechanism, ensuring that the laser beam and the surface of the work roll always maintain the correct relative position. The vibration analysis and suppression system effectively reduces the interference of vibration on the cladding process, and the multi-field joint alignment system dynamically maintains the coaxiality of the work roll axis and the laser processing axis. The coordinated work of these three subsystems constitutes a complete intelligent adaptive control system to cope with various complex disturbance factors in the laser cladding process, thereby ensuring the uniformity of cladding layer thickness and the consistency of surface quality.

[0160] Figure 5 The experimental design employed a normalization method, normalizing seven performance indicators—positioning accuracy, thermal compensation accuracy, vibration amplitude, alignment accuracy, thickness uniformity, surface roughness, and response time—using their respective corresponding comparative benchmark values. This constructed a seven-dimensional normalized data matrix for each sample. Box plots were then drawn along these indicator dimensions to illustrate the data distribution characteristics. In the box plots, the boxes represent the quartile ranges of the data, the median represents the median, the upper and lower bands represent the maximum and minimum value ranges, and circles indicate outliers. Figure 5 As can be seen, the normalized index values ​​of Examples 1-3 are significantly lower than those of the comparative example, with most of the index values ​​concentrated in the range of 0.1 to 0.3, while the index values ​​of the comparative example are scattered in the range of 0.8 to 1.5. Figure 5The red dashed line represents the comparative reference value of 1.0. The fact that the indicator values ​​of the embodiments are significantly lower than this reference line indicates that their performance is significantly better than that of the comparative example. The height of each indicator box reflects the degree of data dispersion; the generally shorter boxes in the embodiments indicate better performance stability. This invention achieves high-precision positioning and stable clamping in the laser cladding repair process of large work rolls through the combined application of innovative technologies such as multi-physics field coupling prediction, frequency-domain and time-domain joint vibration analysis, and multi-field collaborative optimization. This systematic technological innovation brings about a comprehensive performance improvement rather than an improvement in a single indicator.

[0161] Figure 6 The vibration suppression effects of Example 1 and Comparative Example 2 were compared. The experiment used simulated vibration signals to demonstrate the differences between the two schemes. The time vector was set to 0 to 2 seconds, the sampling interval was 0.001 seconds, the main frequency was set to 15Hz (corresponding to the working roller rotation frequency), and the high-frequency vibration was set to 45Hz. Example 1 employed an active-passive hybrid suppression strategy, with a vibration signal amplitude of 0.82 m / s². 2 Comparative Example 2, employing only a passive suppression strategy, exhibits a vibration signal amplitude of 2.35 m / s. 2 .from Figure 6 The time-domain waveforms in subplot (a) clearly show that the vibration signal amplitude in Example 1 is significantly smaller than that in Comparative Example 2. The relatively stable waveform in Example 1 indicates that the vibration is effectively suppressed, while the larger oscillation amplitude in Comparative Example 2 indicates severe vibration. Figure 6 The spectral comparison of subplot (b) shows that the peak values ​​of Comparative Example 2 at 15Hz and 45Hz are significantly higher than those of Example 1, indicating that the vibration energy of Comparative Example 2 at the working roller rotation frequency and its high-frequency components is much greater than that of Example 1. The active-passive hybrid suppression strategy adopted in this invention has significant advantages. The viscoelastic damper in the passive damping section utilizes the energy dissipation of internal friction of the material to suppress vibrations of various frequencies to a certain extent. The air spring has a very low natural frequency and has a good isolation effect on high-frequency vibrations. The magnetorheological damper in the active damping section adapts to different vibration states by adjusting the damping coefficient in real time. The piezoelectric ceramic actuator generates a reverse action force to actively cancel the vibration when it detects vibration. In particular, for high-frequency vibrations caused by laser shock, the extremely fast response speed of the piezoelectric ceramic actuator can track and suppress them in real time. This combination of active and passive methods achieves effective control of broadband vibrations. Compared with Comparative Example 2, which only uses passive damping, the vibration amplitude is reduced by more than 65%.

[0162] Figure 7The dynamic changes in eccentricity over time during the alignment process are shown in Example 1 and Comparative Example 3. The experiment simulated the alignment calibration process starting from an initial eccentricity of 0.08 mm, spanning 0 to 30 seconds. Example 1 employs a multi-field joint alignment mechanism involving thermal gradient field, electromagnetic force field, and mechanical force field, while Comparative Example 3 uses only the traditional mechanical alignment method. The change in eccentricity is simulated by adding a sinusoidal oscillation term to an exponential decay function to represent the characteristics of the actual alignment process. Figure 7 It can be clearly seen that the eccentricity curve of Example 1 decreases at a faster rate, approaching the final accuracy of 0.015mm in about 8 seconds, while the eccentricity curve of Comparative Example 3 decreases slowly, barely approaching the final accuracy of 0.095mm in 30 seconds. Figure 7 The blue dashed line indicates the final alignment accuracy of 0.015mm for Example 1, while the red dashed line indicates the final alignment accuracy of 0.095mm for Comparative Example 3. The difference between the two dashed lines visually demonstrates the significant difference in alignment accuracy between the two schemes. The smaller oscillation amplitude of the curve in Example 1 indicates a smooth alignment process, while the larger oscillation amplitude of the curve in Comparative Example 3 indicates an unstable adjustment process. The multi-field joint alignment method of this invention has significant advantages such as fast alignment speed, high final accuracy, and stable adjustment process. Its working principle is as follows: thermal gradient alignment generates a temperature difference by selectively heating a specific position of the work roller, and uses the thermal expansion effect to drive the axis movement; electromagnetic field alignment achieves non-contact adjustment by coordinating and controlling the current of each electromagnet to generate different attraction forces at different positions of the work roller; and mechanical field alignment provides final precise positioning through a servo electric ball screw. The three physical fields have complementary characteristics: thermal gradient centering is gentle but slow in response, electromagnetic centering is fast but has a limited force range, and mechanical centering is highly accurate but may cause impact. By using a sequential quadratic programming algorithm to globally optimize the control parameters of the three fields, the centering time and energy consumption are minimized while meeting the centering accuracy requirements, thus achieving fast and high-precision centering calibration. Compared with Comparison 3, which only uses mechanical centering, the centering accuracy is improved by more than 87% and the centering speed is significantly accelerated.

[0163] Figure 8 The comparison shows the circumferential distribution of the cladding layer thickness in Example 1 and Comparative Example 1. In the experiment, 360 measurement points were evenly distributed along the circumferential direction (0 to 360 degrees) on the surface of the work roll, and the cladding layer thickness at each point was measured using an ultrasonic thickness gauge. The average thickness was set to 2.5 mm. The standard deviation of the thickness in Example 1, calculated based on its thickness uniformity of 3.2%, was approximately 0.08 mm, while the standard deviation of the thickness in Comparative Example 1, calculated based on its thickness uniformity of 12.5%, was approximately 0.31 mm. Furthermore, the thickness distribution in Comparative Example 1 was superimposed with periodic deviations to simulate systematic deviations caused by positioning instability during actual processing. Figure 8As shown in subplot (a), the measurement points in Example 1 are densely distributed around the average value of 2.5 mm, with a very small dispersion range. Most points are located between 2.4 and 2.6 mm. The average value indicated by the red dashed line matches the actual measurement point height, and the uniformity index of 3.2% marked in the upper right corner indicates that the thickness is very uniform. From Figure 8 As shown in subplot (b), the measurement points in Comparative Example 1 are widely scattered, with data points ranging from 2.2 to 2.8 mm and exhibiting obvious periodic fluctuations, indicating that the cladding layer thickness is highly uneven. The uniformity index of 12.5% ​​marked in the upper right corner confirms this. The comparison of the two subplots clearly demonstrates the significant advantages of the present invention in controlling the uniformity of the cladding layer thickness. The thickness uniformity of Example 1 is improved by more than 74% compared to Comparative Example 1. The multi-physics coupling prediction model used in this invention can accurately predict the thermal deformation trend of the work roll under laser heating. The three-layer cascaded compensation mechanism compensates for thermal deformation in real time, ensuring that the laser beam and the surface of the work roll always maintain a constant working distance. The vibration analysis and suppression system effectively reduces the disturbance of vibration to the cladding process. The multi-field joint alignment system dynamically maintains the coaxiality of the work roll axis and the laser processing axis. The coordinated work of these three subsystems ensures the high stability of the laser cladding process, thereby achieving a high degree of uniformity in the cladding layer thickness. Improving the uniformity of thickness is directly related to the performance and lifespan of the repaired work roll. An uneven cladding layer can lead to localized stress concentration and fatigue cracks, while a highly uniform cladding layer can ensure the consistency and reliability of the overall performance of the work roll.

[0164] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A positioning and clamping method for repairing the surface of large work rolls in steel plants based on laser cladding, characterized in that, Includes the following steps: S1. Establish a multi-physics field coupled prediction model: Arrange multiple measurement sections along the axial direction on the surface of the working roll. Each section is equipped with a temperature sensor, a strain sensor and a displacement sensor to establish a three-field coupled prediction model of temperature field, stress field and displacement field. S2. Perform multi-step look-ahead prediction: Use partial least squares regression algorithm combined with Kalman filter to establish a combined prediction model, and predict the thermal deformation at future moments based on historical data of temperature, stress and displacement. S3. Implement adaptive compensation control: Calculate the compensation control amount based on the predicted thermal deformation amount, and perform compensation through a three-layer cascaded compensation mechanism, including coarse compensation of hydraulically driven V-shaped bracket, fine compensation of piezoelectric ceramic actuator and compensation of laser head Z-axis adjustment mechanism. S4. Perform vibration feature extraction: Collect vibration signals at key locations of the positioning and clamping system, and extract time-domain features, frequency-domain features, and time-frequency-domain features; S5. Perform vibration source classification and suppression: Use support vector machine to classify and identify vibration features, and adopt an active-passive hybrid suppression strategy according to the identified vibration type. S6. Perform axis deviation detection: Use three methods, laser scanning, machine vision and strain field detection, to detect the axis deviation of the work roll; S7. Perform multi-field joint alignment: Based on the detected axis deviation, start the thermal gradient field alignment system, electromagnetic force field alignment system and mechanical force field alignment system, and use the sequential quadratic programming algorithm to coordinately optimize the control parameters of the three fields; S8. Perform laser cladding process control: During the cladding process, perform multiphysics field coupling prediction and compensation, vibration suppression and dynamic centering maintenance in parallel; S1 specifically includes: S11. Configure the sensor array: Arrange the sensor array evenly along the axial direction on the surface of the work roller. Each measurement section is equipped with multiple infrared temperature sensors to collect temperature field data, multiple strain gauge sensors to collect stress field data, and multiple laser displacement sensors to collect displacement field data. S12. Establish a coupled prediction model: Based on the collected temperature field data... Stress field data and displacement field data Establish a three-field coupling relationship, including the temperature field. The stress field is determined by laser power, scanning speed, and substrate temperature. From temperature field The coefficient of thermal expansion and the elastic modulus are determined, and the displacement field is... From stress field The moment of inertia and length of the working roll are determined; S2 specifically includes: S21. Extract latent variables: The temperature of each measurement section... ,stress Displacement Historical data constitutes the input matrix The corresponding thermal deformation amounts are used to form the output matrix. Extracted by partial least squares regression One potential variable; S22. Establishing the state-space model: Constructing the state equations of the Kalman filter. and observation equations ;in for The state vector at any given time includes the temperature, stress, and displacement values ​​of each measured section; for The input vector at each moment includes laser power, scanning speed, and rotational angular velocity; This is the state transition matrix; The input matrix; This is process noise; for The observation vector at time; The observation matrix; To observe noise; S23. Perform multi-step prediction: based on the current time... state vector and input vector The future is calculated recursively using the state equation. Thermal deformation at step time ,in For prediction The amount of thermal deformation at any given time, measured in mm; To predict the number of steps; S4 specifically includes: S41. Vibration signal acquisition: Install acceleration sensors at the chuck, bearing housing, and V-bracket positions to acquire vibration acceleration signals. ;in for The vibration acceleration value at time t, its unit is m / s². 2 ; It is time, and its unit is seconds (s); S42. Extracting temporal features: Calculating the root mean square value Calculate the peak factor ;in This is the root mean square value, and its unit is m / s. 2 ; This represents the number of sampling points; For the first The vibration acceleration value at each sampling point is expressed in m / s². 2 ; This is the summation operator; This is the square root operator; Peak factor; This represents the maximum value of the vibration signal, and its unit is m / s. 2 ; S43. Extracting frequency domain features: For vibration signals... The spectrum is obtained by performing a fast Fourier transform. Identify main frequency Calculate the spectral entropy ;in For frequency domain spectrum; Frequency (Hz); The main frequency is the frequency corresponding to the maximum amplitude of the spectrum, and its unit is Hz; For spectral entropy; For frequency Normalized energy probability at the location; For logarithmic operators; S44. Extracting time-frequency domain features: For vibration signals... conduct Layer wavelet packet decomposition, calculating the th Layer energy ;in The number of wavelet packet decomposition layers; For the first Layer energy characteristics; For the first Layer Wavelet coefficients; S45. Constructing Feature Vectors: Combining time-domain features, frequency-domain features, and time-frequency-domain features to form feature vectors. Used for classifying vibration sources; where RMS is the root mean square value, reflecting the energy of the vibration signal, the larger the value, the more intense the vibration; CF is the peak factor, reflecting the waveform characteristics of the vibration signal, dimensionless, distinguishing different types of vibration, periodic vibration has a small peak factor, while impact vibration has a large peak factor. The main frequency is the frequency corresponding to the maximum spectral amplitude, reflecting the main frequency components of the vibration. Different vibration sources have specific frequency characteristics. S5 specifically includes: S51. Classifying vibration sources: This involves classifying the feature vectors... Input the support vector machine classifier to identify the vibration type as unbalanced vibration, bearing failure vibration, laser shock vibration, or resonant vibration. S52. Suppressing Unbalanced Vibration: Unbalanced vibration is passively suppressed by adjusting the position of the counterweight, and actively suppressed by adjusting the speed of the servo motor. The speed compensation algorithm is as follows: ;in for The angular velocity after compensation at any given time is expressed in rad / s. To set the angular velocity, its unit is rad / s; This is the unbalanced compensation coefficient; The amplitude of the vibration is expressed in m / s. 2 ; It is a sine function; This is the rotational angular velocity, and its unit is rad / s; This is the phase angle, and its unit is rad; S53. Suppressing Bearing Vibration: For bearing vibration, passive suppression is achieved by adding a viscoelastic damper, and active suppression is achieved by adjusting the damping coefficient using a magnetorheological damper. Damping force... ;in for Damping force (N) at time t; For the current The varying damping coefficient, with units of N·s / m; This is the control current of the magnetorheological damper, and its unit is A; for The vibration velocity at any given moment, its unit is m / s; S54. Suppressing Laser Shock Vibration: For laser shock vibration, passive suppression is achieved through an air spring vibration isolation platform, and active suppression is achieved by generating a reverse force through a piezoelectric ceramic actuator. The control law is as follows: ;in for The constant output force of the actuator, measured in N; The transfer function of the adaptive filter; For the Laplace operator; for The disturbance force at any given time is measured in N.

2. The method according to claim 1, characterized in that, S3 specifically includes: S31. Calculate the compensation amount: Based on the predicted thermal deformation amount... Through the compensation control law Calculate radial compensation displacement ;in This represents the radial compensation displacement, and its unit is mm. For proportional gain; For prediction The amount of thermal deformation at any given time, measured in mm; This is the differential gain; For prediction The amount of thermal deformation at any given time, measured in mm; S32. Adjust control gain: Based on the predicted thermal deformation. The magnitude and rate of change are adaptively adjusted by fuzzy rules to adjust the proportional gain. and differential gain ; S33. Implement three-tier compensation: adjust the compensation amount. The compensation mechanism is distributed to three layers. It performs millimeter-level coarse compensation by driving the V-shaped bracket through the hydraulic system, micron-level fine compensation by the piezoelectric ceramic actuator, and submicron-level ultra-fine compensation by the laser head Z-axis adjustment mechanism.

3. The method according to claim 1, characterized in that, S6 specifically includes: S61. Perform laser scanning inspection: Install laser profile scanners at both ends of the work roll to collect the profile point set at the roll end during the rotation of the work roll. The coordinates of the circle center were fitted using the least squares method. Calculate the eccentricity and eccentricity ;in For the first The coordinates of the contour points are in mm. Number of data collection points; The coordinates of the fitted circle's center are in mm. The eccentricity measured by laser scanning is expressed in mm. The coordinates of the reference axis are in mm. The eccentricity angle is measured by laser scanning, and its unit is rad. It is the arctangent function; S62. Perform machine vision inspection: Capture images of the roller end face using an industrial camera, identify the center coordinates of the circle using edge detection and circle recognition algorithms, and calculate the eccentricity. and eccentricity ;in The eccentricity measured by machine vision is expressed in mm. The eccentricity angle is measured by machine vision and its unit is rad. S63. Perform strain field testing: Collect strain distribution data at the V-shaped bracket contact point and calculate the eccentric load index. Determine whether eccentricity exists; among which This is the off-center load index; This represents the maximum strain value. This is the minimum strain value; S64. Fusion Detection Results: The detection results from laser scanning, machine vision, and strain field methods are weighted and fused to obtain the final eccentricity. and eccentricity ;in The eccentricity after fusion is expressed in mm. The weights for the laser scanning method; The weights for machine vision methods; The eccentricity angle after fusion is expressed in rad.

4. The method according to claim 3, characterized in that, Specifically, S7 and S8 include: S71. Start thermal gradient alignment: The annular electric heaters at both ends of the work roll heat the rollers in sections according to the eccentric direction, generating a temperature difference. thermally induced deformation ;in This represents the amount of thermally induced deformation, and its unit is mm. is the coefficient of linear expansion, and its unit is 1 / ℃; This refers to the length of the work roll, and its unit is mm. This refers to the temperature difference, and its unit is °C. This refers to the diameter of the work roll, which is expressed in mm. S72. Activate electromagnetic field alignment: Electromagnetic attraction is generated through the electromagnet array on both sides of the V-shaped bracket. ;in This refers to electromagnetic attraction, and its unit is N; This refers to the number of coil turns. Electric current is measured in amperes (A). ρ is the permeability of free space, and its unit is H / m; Relative permeability; The area of ​​the magnetic poles is measured in meters (m²). 2 ; This refers to the air gap, and its unit is meters (m). S73. Start the mechanical field centering: The position of the V-shaped bracket is adjusted by driving the V-shaped bracket through a servo electric ball screw, and the position closed-loop control is adopted. ;in for The control quantity at any given time is measured in millimeters (mm). For proportional gain; for The eccentricity at any given time, measured in mm; This is the integral gain; This is the differential gain; for The eccentricity at any given time, measured in mm; S74. Perform multi-field collaborative optimization: Establish the optimization objective function. The optimal heating power is solved using a sequential quadratic programming algorithm. Electromagnetic current and mechanical displacement ;in To optimize the objective function; This is the eccentricity weighting coefficient; This refers to the eccentricity, and its unit is mm. This is the thermal field weighting coefficient; Heating power, measured in watts (W). This is the electromagnetic field weighting coefficient; For mechanical field weighting coefficients; In S8, the following are performed simultaneously during the laser cladding process: predicting thermal deformation through a multi-physics field coupling prediction model and compensating in real time through a three-layer compensation mechanism; identifying vibration types through a vibration analysis model and suppressing vibration through an active-passive hybrid approach; and monitoring axis deviation through a multi-modal detection system and maintaining alignment accuracy through a multi-field joint system.

5. A positioning and clamping system for implementing the method described in any one of claims 1-4, based on laser cladding repair of the surface of large work rolls in steel plants, characterized in that, include: The sensing system includes an array of temperature sensors, an array of strain sensors, and an array of displacement sensors arranged along the axial direction of the work roll, which collects temperature field, stress field, and displacement field data. It includes vibration sensors installed at key positions in the positioning and clamping system to collect vibration signals; laser profile scanners and industrial cameras installed at both ends of the work roller to detect axial deviation; temperature sensor arrays, strain sensor arrays, displacement sensor arrays and vibration sensors are connected to the control system through a data acquisition card, and laser profile scanners and industrial cameras are connected to the control system through a communication interface; The execution system includes a three-layer cascaded compensation mechanism for thermal deformation compensation, an active-passive hybrid vibration damping mechanism for vibration suppression, and a multi-field joint alignment mechanism for alignment calibration. The three-layer cascaded compensation mechanism includes a hydraulically driven V-shaped bracket, a piezoelectric ceramic actuator mounted on the V-shaped bracket, and a Z-axis adjustment mechanism connected to the laser head. The active-passive hybrid vibration damping mechanism includes a viscoelastic damper and an air spring as passive vibration damping components, and a magnetorheological damper and a piezoelectric ceramic actuator array as active vibration damping components. The multi-field joint alignment mechanism includes annular electric heaters inside the chucks at both ends of the work roll, an electromagnet array on both sides of the V-shaped bracket, and a servo-driven electric ball screw that drives the V-shaped bracket. The hydraulic drive system, piezoelectric ceramic actuator, Z-axis adjustment mechanism, magnetorheological damper, piezoelectric ceramic actuator array, annular electric heater, electromagnet array, and servo-driven electric ball screw are all connected to the control system via actuators. The control system includes a data acquisition and processing module, a predictive compensation control module, a vibration analysis and suppression control module, and a centering optimization control module. The data acquisition and processing module receives sensor data through a data acquisition card, performs preprocessing, and then transmits the data to the predictive compensation control module, the vibration analysis and suppression control module, and the centering optimization control module. The predictive compensation control module includes a partial least squares regression unit, a Kalman filter unit, and a compensation control unit. It establishes a multi-physics coupled prediction model, performs multi-step look-ahead prediction, and generates compensation control commands to be sent to the three-layer cascaded compensation mechanism. The vibration analysis and suppression control module includes a feature extraction unit, a support vector machine classification unit, and a vibration control unit. It extracts vibration features, identifies vibration types, and generates suppression control commands to be sent to the active-passive hybrid vibration reduction mechanism. The centering optimization control module includes a deviation detection unit, a sequential quadratic programming optimization unit, and a multi-field control unit. It integrates multi-modal detection results, optimizes multi-field control parameters, and generates centering control commands to be sent to the multi-field joint centering mechanism.

6. The system according to claim 5, characterized in that, The temperature sensor array includes an infrared temperature sensor and a thermal imaging camera; the infrared temperature sensor is divided into multiple measurement sections along the working roller axis, and multiple infrared temperature sensors in each measurement section are evenly distributed circumferentially; the strain sensor array includes strain gauge sensors and fiber optic strain sensors, with the strain gauge sensors and infrared temperature sensors arranged in a staggered manner; the displacement sensor array includes laser displacement sensors, which are arranged opposite each other in each measurement section; the vibration sensor includes triaxial acceleration sensors, which are respectively mounted on the chuck, bearing housing, and V-shaped bracket; In the three-layer cascaded compensation mechanism, the hydraulically driven V-shaped bracket is connected to the fixed column via a hydraulic cylinder, which is controlled by a hydraulic control valve; the piezoelectric ceramic actuator is installed at the contact position between the V-shaped bracket and the work roller, with one end of the piezoelectric ceramic actuator fixedly connected to the V-shaped bracket and the other end in contact with the surface of the work roller; the Z-axis adjustment mechanism is connected to the laser head mounting bracket via a guide rail, and the drive end of the Z-axis adjustment mechanism is connected to the output shaft of the voice coil motor; the control commands for the three-layer compensation mechanism are output from the compensation control unit of the predictive compensation control module. In the active-passive hybrid vibration reduction mechanism, a viscoelastic damper is installed between the bearing housing and the foundation; an air spring is installed between the V-shaped bracket support base and the foundation; a magnetorheological damper is installed between the bearing housing and the foundation, arranged in parallel with the viscoelastic damper, and the current input terminal of the magnetorheological damper is connected to the PWM controller; a piezoelectric ceramic actuator array is installed on the V-shaped bracket support base, and the drive voltage input terminal of the piezoelectric ceramic actuator is connected to the piezoelectric driver; the PWM controller and the piezoelectric driver receive control signals output from the vibration control unit of the vibration analysis and suppression control module; In the multi-field joint centering mechanism, an annular electric heater is embedded in the inner wall of the chucks at both ends of the work roll. The annular electric heater is divided into multiple independent control zones, each of which is independently powered by a heating controller. An electromagnet array is mounted on the brackets on both sides of the V-shaped bracket, with the magnetic pole faces of the electromagnets facing the surface of the work roll. The coils of the electromagnets are powered by an electromagnetic driver. The nut end of the servo electric ball screw is connected to the movable column, and the screw end is connected to the output shaft of the servo motor. The movable column is connected to the fixed base via a guide rail. The heating controller, electromagnetic driver, and servo motor driver receive control signals output from the multi-field control unit of the centering optimization control module. The hardware architecture of the control system includes an industrial computer, a motion control card, a data acquisition card, and a fieldbus. The industrial computer runs a real-time operating system and connects to the motion control card and the data acquisition card through an expansion interface. The motion control card connects to hydraulic control valve drivers, piezoelectric drivers, voice coil motor drivers, PWM controllers, heating controllers, electromagnetic drivers, and servo motor drivers. The data acquisition card connects to temperature sensors, strain sensors, displacement sensors, and vibration sensors. The laser profile scanner and industrial camera communicate with the industrial computer through the fieldbus.

Images