High-precision dynamic calibration algorithm for industrial CT

Abstract

The application discloses a kind of high-precision dynamic calibration algorithms of industrial CT, it is related to the field of industrial CT technology, algorithm includes: the state space model including kinematics factor (machining error, assembly error etc.) and dynamics factor (mass, friction etc.) is constructed, real-time correction is realized using model predictive control (MPC) combined with dynamic compensation algorithm (disturbance observation, end effect compensation, friction fitting), data is collected using laser interferometer (resolution ≤1nm) and grating encoder (accuracy ±2 μm), and the parameters are optimized by PSNR evaluation and Simulink simulation.Compared with traditional PID algorithm, MPC supports multivariable optimization, nonlinear model adaptation and constraint embedding, significantly improves dynamic positioning accuracy (micron level), response speed and image quality (PSNR ≥45dB), suitable for dynamic detection of coal, rock and other materials, effectively reduces artifact and measurement error, and provides an efficient and reliable dynamic calibration scheme for industrial CT system.

Description

Technical Field

[0001] This invention relates to the field of industrial CT technology, specifically a high-precision dynamic calibration algorithm for industrial CT. Background Technology

[0002] Existing industrial CT (Computed Tomography) systems, when performing dynamic scans, suffer from mechanical defects, such as kinematic factors—machining errors, mechanical tolerances, assembly errors, reducer precision, reducer idle distance, and grating ruler engraving errors; or dynamic factors—mass, inertia tensor, friction, and flexible transmission—leading to significant errors between actual theoretical and measured values. These defects result in the following problems: Image blurriness: The accuracy of the motion system determines the spatial resolution of CT images. If the accuracy of the motion system is insufficient, it will result in unclear images, shadows, and make it very difficult to identify the internal structure of objects.

[0003] Measurement error: Poor motion precision may lead to positional deviations during the scanning process, thus affecting the accuracy of the measurement results. This error may manifest as inaccurate dimensional measurements, incorrect positioning, etc. For example, in SODSDD calculations, it is difficult to accurately evaluate the object being measured.

[0004] Reconstruction Issues: In CT scans, data reconstruction relies on accurate scan data. Poor motion precision can lead to inaccurate data acquisition, affecting the quality of the reconstruction. This can result in distortion, artifacts, and other problems in the reconstructed image, reducing image quality.

[0005] Impact Analysis: In industrial CT scans, the scan results are typically analyzed to assess the internal structure and defects of an object. Poor motion accuracy can lead to inaccurate analysis results and even mislead the analyst's judgment.

[0006] In view of the above, this application is hereby submitted. Summary of the Invention

[0007] The purpose of this invention is to provide a high-precision dynamic calibration algorithm for industrial CT to solve the problems mentioned in the background art.

[0008] To solve the above-mentioned technical problems, the present invention provides a high-precision dynamic calibration algorithm for industrial CT, comprising the following steps: Step 1: Based on the dynamic characteristics and kinematic constraints of the industrial CT motion control system, construct a state-space model that includes kinematic and dynamic factors. The state-space model includes the dynamic equations of the mechanical system, the state equations of the electrical system, and the state equations of the control system. Step 2: The Model Predictive Control (MPC) algorithm is adopted to predict the state sequence and input sequence of the system in the prediction time domain based on the discretized state-space model. The calibration parameters are fitted by polynomial through a high-order fitting algorithm, and the motion platform is corrected in real time by a dynamic compensation algorithm. Step 3: Use high-precision sensors to collect command position data and actual position data of the CT control system in real time. After filtering and noise reduction preprocessing, the data is used for model verification and parameter optimization. Step 4: Compare the experimental data with the model prediction results, adjust the parameters of the high-order fitting algorithm to approximate the real dynamic characteristics of the system, and verify the calibration accuracy through third-party sensors and CT scan results. Through step-by-step model building, prediction compensation, data acquisition and iterative optimization, the positioning accuracy and response speed of industrial CT dynamic scanning are systematically improved, effectively compensating for mechanical errors and dynamic interference. It is suitable for dynamic internal structure detection of materials such as coal and rock, and significantly improves the clarity of CT images and detection reliability.

[0009] Furthermore, in step one, kinematic factors include mechanical tolerances, machining errors, assembly errors, and reducer precision errors, while dynamic factors include mass, inertia tensor, friction, and external disturbances. Clarifying the specific types of kinematic and dynamic factors makes the model construction more comprehensive and accurate, enabling a more precise description of system characteristics, providing a reliable foundation for subsequent control and compensation, and improving the algorithm's adaptability to complex industrial environments.

[0010] Furthermore, in step one, the model uses the zero-order hold method to model the mechanical system and the forward discrete Euler method to model the electrical system. Using the zero-order hold method and the forward discrete Euler method to discretize the mechanical system and the electrical system respectively can accurately convert the continuous system into a discrete system, preserve the dynamic characteristics of the system, improve the prediction accuracy of the model, and provide a more accurate mathematical model for subsequent control and compensation.

[0011] Furthermore, in step two, the dynamic compensation algorithm includes designing an extended state observer to estimate disturbances and friction, using an IP controller to compensate for the end effect error of the linear motor, and modeling the friction curve using a Gaussian fitting method. Through the combination of multiple dynamic compensation strategies, the influence of external disturbances and system nonlinear factors on CT scan accuracy is effectively suppressed, the anti-interference ability and dynamic response performance of the system are improved, the positioning error and jitter during the motion process are reduced, and the image reconstruction quality is improved.

[0012] Furthermore, in step two, the objective function of model predictive control (MPC) dynamically adjusts the weight parameters of displacement, velocity, and current according to the scanning scenario. The displacement weight is increased in spiral CT / bias scanning scenarios, the velocity weight is increased in ordinary CT / DR scanning scenarios, and the current weight is increased in low-heat scenarios. By dynamically adjusting the weight parameters of the MPC objective function according to different scanning scenarios, the algorithm can adapt to different scanning requirements. While ensuring scanning accuracy, it optimizes the system's energy consumption and heat generation, extends the equipment's service life, and improves the system's adaptability and flexibility.

[0013] Furthermore, in step three, the high-precision sensor includes a laser interferometer with a resolution of not less than 1 nm and a linear phase grating encoder with a grating pitch of not more than 8 μm and an accuracy level of ±2 μm. The use of a high-resolution laser interferometer and a high-precision linear phase grating encoder for data acquisition ensures the accuracy and real-time performance of position measurement, providing reliable data support for subsequent model verification and parameter optimization, thereby further improving the accuracy and reliability of the calibration algorithm.

[0014] Furthermore, in step four, the image reconstruction quality is evaluated using peak signal-to-noise ratio (PSNR), and the weight parameters are optimized based on Simulink simulation. Evaluating the image reconstruction quality using PSNR and verifying the system response performance using Simulink simulation can intuitively assess the effectiveness of the calibration algorithm, promptly identify and resolve problems, optimize the weight parameters, further improve the quality of CT images and the performance of the system, and ensure the effectiveness and reliability of the algorithm.

[0015] The high-precision dynamic calibration algorithm of industrial CT is used to improve the positioning accuracy, response speed and positioning stability during dynamic scanning, meeting the internal structure inspection needs of coal, rock, lithium battery, CPU, PCB board, composite material, metal material and other materials under dynamic loading conditions. The algorithm is integrated into the industrial CT system, realizing a comprehensive improvement in positioning accuracy, response speed and stability during dynamic scanning, providing an efficient and reliable technical solution for dynamic performance testing of materials and industrial non-destructive testing, and has significant engineering application value.

[0016] Compared with the prior art, the beneficial effects of the present invention are: 1. Improved Image Clarity: The precision of the motion system determines the spatial resolution of CT images. Improving motion precision can significantly reduce positional deviations and jitter during the scanning process, thereby ensuring more accurate data acquisition. This will help generate clearer and more detailed images, enabling inspectors to observe the object being measured more accurately.

[0017] 2. Improved Measurement Accuracy: Precise measurement is paramount in industrial CT scanning. Improving motion accuracy reduces measurement errors, making dimensional measurements and positional positioning more accurate. This is crucial for assessing an object's internal structure, identifying defects, and implementing quality control.

[0018] 3. Optimization of Data Reconstruction Results: Data reconstruction is a crucial step in industrial CT scanning. Improving motion accuracy ensures more accurate and complete acquired data, thereby enhancing the effectiveness of data reconstruction. This will help generate more realistic and accurate 3D images, providing strong support for subsequent analysis and evaluation.

[0019] 4. Enhanced Analytical Accuracy: In industrial CT scans, the scan results are typically analyzed to assess an object's internal structure, defects, etc. Improving motion precision enhances the accuracy of CT scans, enabling engineers to more accurately determine an object's internal structure and defects. This, in turn, can lead to improved product quality, increased production efficiency, and reduced production costs.

[0020] 5. Reduce artifacts and distortion: Poor motion accuracy can lead to artifacts and distortion during scanning. Improving motion accuracy can reduce these problems, resulting in more realistic and reliable images and data. Attached Figure Description

[0021] Figure 1 A schematic diagram of linear measurement using a laser interferometer in a high-precision dynamic calibration algorithm for industrial CT. Figure 2 This is a schematic diagram of feedback control consisting of an observer, external disturbance, and the model of the object being measured in a high-precision dynamic calibration algorithm for industrial CT. Figure 3 This is a schematic diagram of the dynamic calibration process, which consists of mechanical errors, dynamic calibration commands, and compensators, in the high-precision dynamic calibration algorithm for industrial CT. Figure 4 A comparison chart of friction force feedback values ​​and fitted values ​​in a high-precision dynamic calibration algorithm for industrial CT. Detailed Implementation

[0022] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] Please see Figures 1-4This invention provides a technical solution: a high-precision dynamic calibration algorithm for industrial CT. First, a mathematical model of the system is established based on the dynamic characteristics and kinematic constraints of the motion control system. Then, the MPC algorithm is used to predict the future behavior of the system, and a high-order fitting algorithm is used to perform polynomial fitting on the calibration parameters. This is combined with model prediction and compensation algorithms to achieve an accurate description of the system's dynamic characteristics.

[0024] A dynamic compensation algorithm is introduced to perform real-time correction on the motion platform of the industrial CT system, ensuring the accuracy of measurement results. The experiments cover motion control under different motion axes, positions, and directions, guaranteeing the reliability and repeatability of the experimental results.

[0025] During the dynamic calibration experiment, a high-precision sensor (laser interferometer) was used to acquire real-time data on the actual and commanded positions of the CT control system. The acquired data underwent preprocessing steps such as filtering and noise reduction before being used for subsequent comparison and calculation.

[0026] The experimental data were compared and analyzed with the simulation results predicted by the calibration algorithm to evaluate the accuracy and applicability of the model. Based on the analysis results, the higher-order fitting algorithm was further optimized and adjusted to improve the calibration accuracy. Simultaneously, through an iterative optimization process, the model gradually approximates the true dynamic characteristics of the system.

[0027] The model parameters and calibration compensation values ​​were applied to the industrial CT control system, and the results were verified by a third-party high-precision sensor and actual industrial CT scan results.

[0028] Based on the test results, the control system will be further debugged and optimized to ensure the accuracy and reliability of the industrial CT system in detecting changes in the internal structure of the sample during actual scanning and inspection.

[0029] This achieved the following: 1. Improved positioning accuracy and response speed of industrial CT systems in both static and dynamic environments, enabling precise measurement of the tested samples. 2. Enhanced stability and robustness of industrial CT systems; the application of dynamic compensation algorithms ensures stable measurement performance under complex conditions, reducing sensitivity to external interference. 3. Provided reliable technical support for optimizing industrial CT scan results and image reconstruction algorithms. 4. Improved the efficiency of data acquisition and processing in industrial CT scans, shortening the testing cycle.

[0030] Example 1: The high-precision dynamic calibration algorithm for industrial CT scanners includes the following steps: Step 1: Based on the dynamic characteristics and kinematic constraints of the industrial CT motion control system, a state-space model incorporating kinematic and dynamic factors is constructed. This model includes the dynamic equations of the mechanical system, the state equations of the electrical system, and the state equations of the control system. Kinematic factors include mechanical tolerances, machining errors, assembly errors, and reducer precision errors, while dynamic factors include mass, inertia tensor, friction, and external disturbances. Clearly defining the specific types of kinematic and dynamic factors ensures a more comprehensive and accurate model, enabling a more precise description of system characteristics and providing a reliable foundation for subsequent control and compensation. This also improves the algorithm's adaptability to complex industrial environments. The model employs the zero-order hold method for modeling the mechanical system and the forward discrete Euler method for modeling the electrical system. Discretizing the mechanical and electrical systems using the zero-order hold method and the forward discrete Euler method accurately converts continuous systems into discrete systems, preserving the system's dynamic characteristics and improving the model's prediction accuracy. This provides a more accurate mathematical model for subsequent control and compensation. Step 2: Model Predictive Control (MPC) algorithm is adopted. Based on the discretized state-space model, the system's state sequence and input sequence in the prediction time domain are predicted. A high-order fitting algorithm is used to perform polynomial fitting on the calibration parameters, combined with a dynamic compensation algorithm to correct the motion platform in real time. The dynamic compensation algorithm includes designing an extended state observer to estimate disturbances and friction, using an IP controller to compensate for the linear motor end-effect error, and modeling the friction curve using a Gaussian fitting method. Through the combination of multiple dynamic compensation strategies, the influence of external disturbances and system nonlinear factors on CT scan accuracy is effectively suppressed, improving the system's anti-interference capability and dynamic response performance, reducing positioning errors and jitter during motion, and improving image reconstruction quality. The objective function of Model Predictive Control (MPC) dynamically adjusts the weight parameters of displacement, velocity, and current according to the scanning scenario. The displacement weight is increased in spiral CT / offset scanning scenarios, the velocity weight is increased in ordinary CT / DR scanning scenarios, and the current weight is increased in low-heat scenarios. By dynamically adjusting the weight parameters of the MPC objective function according to different scanning scenarios, the algorithm can adapt to different scanning requirements, optimizing system energy consumption and heat generation while ensuring scanning accuracy, extending equipment lifespan, and improving system adaptability and flexibility.

[0031] Step 3: Real-time acquisition of command position data and actual position data from the CT control system using high-precision sensors. After filtering and noise reduction preprocessing, the data is used for model verification and parameter optimization. The high-precision sensors include a laser interferometer with a resolution of no less than 1 nm and a linear phase grating encoder with a grating pitch of no more than 8 μm and an accuracy level of ±2 μm. The use of a high-resolution laser interferometer and a high-precision linear phase grating encoder for data acquisition ensures the accuracy and real-time performance of position measurements, providing reliable data support for subsequent model verification and parameter optimization, thereby further improving the accuracy and reliability of the calibration algorithm.

[0032] Step 4: Compare the experimental data with the model prediction results, adjust the parameters of the high-order fitting algorithm to approximate the true dynamic characteristics of the system, and verify the calibration accuracy through third-party sensors and CT scan results; evaluate the image reconstruction quality through peak signal-to-noise ratio (PSNR), and optimize the weight parameters based on Simulink simulation; evaluating the image reconstruction quality through PSNR and verifying the system response performance through Simulink simulation can intuitively evaluate the effect of the calibration algorithm, identify and solve problems in a timely manner, optimize the weight parameters, further improve the quality of CT images and the performance of the system, and ensure the effectiveness and reliability of the algorithm; through step-by-step model building, prediction compensation, data acquisition and iterative optimization, the positioning accuracy and response speed of industrial CT dynamic scanning are systematically improved, effectively compensating for kinematic factors and dynamic interference, and are suitable for dynamic internal structure detection of coal, rocks, lithium batteries, CPUs, PCB boards, composite materials, metal materials, etc., significantly improving the clarity of CT images and detection reliability.

[0033] The high-precision dynamic calibration algorithm of industrial CT is used to improve the positioning accuracy, response speed and positioning stability during dynamic scanning, meeting the internal structure inspection needs of coal, rock, lithium battery, CPU, PCB board, composite material, metal material and other materials under dynamic loading conditions. The algorithm is integrated into the industrial CT system, realizing a comprehensive improvement in positioning accuracy, response speed and stability during dynamic scanning, providing an efficient and reliable technical solution for dynamic performance testing of materials and industrial non-destructive testing, and has significant engineering application value.

[0034] It should be noted here that: MPC Algorithm Derivation 1. Establishment of the dynamic model of the motion axis: m + f f( )+ f d( t )= F t; in: xLinear displacement; m : Load quality; f f: Frictional force (which may include static friction, Coulomb friction, and viscous friction); fd External disturbances (such as load changes); Ft Electromagnetic thrust.

[0035] Ft=k * iq ; k Thrust constant; iq q-axis current (linearly related to thrust).

[0036] 2. State-space representation Define state variable x=[x, ,id,iq]^T, Input u=[vd,vq]^T, The output is the displacement x. The state equation is: X˙ =[ ] ; 3. Establish a discrete model Mechanical system: Using the zero-order hold method, Assuming the thrust Ft and friction force during the sampling period Ts Internal constant: X (k+1)=x(k)+Ts* (k)+Ts 2 / 2m (k*iq(k)-Ff); (k+1)=x˙(k)+Ts / m (k*iq(k)-Ff).

[0037] Electrical system: Using forward discrete Euler: Id ( k +1)= id ( k )+ Ts / Ld ( Vd ( k )- Rs * id ( k )+ We ( k ) Lqiq ( k )), Iq(K+1)= iq(k)+Ts / Lq(Vq(k)-Rsiq(k)-We(k)Ldid(k)-We(k) ) .

[0038] 4. Model Prediction Extension Predict the state sequence X(k) and input sequence U(k) in the time domain Np. Includes mechanical and electrical status: X(k) = Ax(k) + BU(k) + D [ ] ; A and B are constructed recursively from the discretized model. D includes the cumulative impact of external disturbances.

[0039] 5. Objective function design Simultaneous optimization of displacement tracking, thrust fluctuation, and current constraint is required: J = + ; Where Qx, Qv, and Qi are the weights of displacement, velocity, and current, Δ v ( k )= v ( k )- v ( k -1) Used to smooth voltage input.

[0040] The physical meaning of the weight matrix: Qx: Weight of displacement tracking error, reflecting the requirements for positioning accuracy.

[0041] Qv: Weight of velocity error, affecting motion smoothness and disturbance resistance.

[0042] Qi: Weight of current (thrust) error, associated with energy efficiency and motor heating.

[0043] Based on the principle of control target priority CT spiral scan / offset scan scenarios: Qx needs to be much larger than Qv and Qi (Qx = Qv= (Qi=1), to ensure that the position tracking error is minimized.

[0044] In typical CT or DR scan scenarios, Increase Qv to suppress velocity fluctuations (e.g., Qv = , Qx= ), Scenarios requiring low fever: Increase Qi to limit the current amplitude (e.g., Qi=10, Qx= ).

[0045] Normalize state and control variables to avoid numerical problems caused by differences in dimensions: =

[0046] =

[0047] = .

[0048] The initial weight values ​​can be set as follows: Qx=

[0049] Qv=

[0050] Qi= .

[0051] Dynamic adjustment strategy: Iterative parameter tuning method: 1. Initially set Qx=1, Qv=0.1, Qi=0.01; 2. Gradually increase Qx until the position error meets the standard; 3. Adjust Qv to suppress speed overshoot; 4. Finally, adjust the Qi to balance the current consumption.

[0052] Frequency domain analysis: By adjusting the weights using the open-loop frequency response function, the CT scanning system can be kept within a tracking bandwidth of 10Hz or higher.

[0053] Special considerations for industrial CT scenarios (1) Scan trajectory characteristics Standard CT or DR scan mode: During the constant speed-dwell phase, extremely high Qx is required to ensure positioning accuracy; during the constant speed scanning phase, Qv is increased.

[0054] Spiral CT scan mode: Qx and Qv need to be balanced, and the ratio of Qx:Qv should be adjusted to approximately 10:1.

[0055] (2) Disturbance resistance requirements Vibration suppression: Increase Qv to enhance damping (Qv=2Qx).

[0056] Load mutation: Appropriately reduce Qi to allow instantaneous overcurrent (Qi reduced by 50%).

[0057] (3) Influence of motor parameters Thrust constant Kf: Qi should be related to kf 2 Inversely proportional, using a high-thrust motor.

[0058] Mechanical resonant frequency Fres: If Fres is close to the control bandwidth, Qv needs to be increased to suppress resonance.

[0059] Steps for determining weights in actual engineering 1. Simulation verification: Build a model in Simulink and test the step / sinusoidal response; Observe the overshoot, settling time, and peak current, and adjust the weights accordingly.

[0060] 2. Experimental calibration: With Qi=1 fixed, scan the (Qx,Qv) combination; Choose the parameter that minimizes the position RMSE.

[0061] 3. Online adaptive: Design rules: Position error Qx( k )= Qx+ α | ex ( k )∣; ex This refers to the positional error; The weights are dynamically adjusted to adapt to different scanning stages.

[0062] The following are recommended values:

[0063] 6. Adjustment of constraints Electrical constraints Current amplitude limit: √(i 2 d + i 2 q) ≤ Imax; Voltage amplitude limit: √ (v) 2 d+ v 2 q) ≤ Vmax; Mechanical constraints; Speed ​​Limit: | | ≤ Vmax; Displacement range: Xmin ≤ X ≤ Xmax; Thrust limit: |Ft| ≤ |Fmax|.

[0064] 7. Special Issues Compilation The first is the end effect of linear motors. An end effect will occur at the end of the stator of a linear motor because there will be tiny gaps at the stator splice. The magnetic flux distribution at this point is different from that in the middle part. Not only is the magnetic field weak, but distortion will also occur. Current fluctuations will occur at this point. An IP controller can be designed to compensate for the current fluctuations here.

[0065] End-of-pipe compensation effect: F t = kf *iq + Δ F end( x ); like Figure 4 As shown, the second method is friction compensation (Gaussian fitting method). The friction force curve fitting method is linear fitting. The dependent variable is the command torque, and the independent variable is the command velocity, which is related to the direction of the command velocity.

[0066] myfittype = fittype('b*sign(x) 'dependent' , {'y'}," independent',{'x'}.... 'coefficients',{'b'}) myfit = fit(cmdvel, cmdTog, myfittype, "startPoint', 0) fitFriction: (myfit, b+0.000)*sign(cmdVel) lab = sprintf( '%0.3f*sign(qdot)', myfit.b): 8. Adjustment of the control framework Thrust-current decoupling: Maximizing the thrust / current ratio by controlling Id=0; Disturbance Observer: Design an extended state observer (ESO) to estimate Fd and friction force in real time; Parameter adaptation: Online identification of m and kf to cope with load changes.

[0067] Basis for comparing experimental data: Comparison method: 1. Discuss using the PSNR (Peak Signal-to-Noise Ratio) method. Principle: The degree of distortion is measured by pixel-level error.

[0068] Advantages: Simple calculation, suitable for rapid evaluation.

[0069] Disadvantages: It has low correlation with human visual perception and cannot distinguish between structural distortion and noise.

[0070] Application scenario: Preliminary evaluation of image compression and denoising algorithms.

[0071] Image differences corresponding to the value: The typical peak signal-to-noise ratio (PSNR) in image and video compression is between 30dB and 50dB, with higher being better.

[0072] A PSNR close to 50dB means that the compressed image has only a very small error.

[0073] When the PSNR is greater than 30dB, it is difficult for the human eye to perceive the difference between the compressed image and the original image.

[0074] When the PSNR is between 20dB and 30dB, the human eye can perceive the difference in the image.

[0075] With a PSNR between 10dB and 20dB, the human eye can still discern the original structure of the image and intuitively judge that there is not much difference between the two images.

[0076] When the PSNR is below 10dB, it is difficult for humans to visually determine whether two images are identical or whether one image is a compressed version of the other.

[0077] Because the MPC algorithm, which combines image model and motion model, is used, each discrete motion point is within a pixel block, resulting in almost no pixel error and a perfect signal-to-noise ratio.

[0078] The positioning sensor uses HEIDENHAIN SUPRADU with a steel substrate.

[0079] Linear phase grating: Pitch: 8µm.

[0080] Linear expansion coefficient: 10*10^(-6) K^(-1).

[0081] Diagnostic interface: 1VPP analog subdivision; Signal period: 4µm; Cutoff frequency: ≥250kHz; Accuracy class: ±2um.

[0082] The laser interferometer used for measurement and calibration is a Renishaw laser interferometer. Laser resolution 1nm.

[0083] like Figure 1 As shown: Higher precision measuring instruments and verification CT equipment are used to improve the overall confidence level of the equipment.

[0084] In summary: as shown below:

[0085] Currently, the commonly used algorithms on the market are traditional PID algorithms. Compared to traditional PID algorithms, MPC natively supports multivariable systems (MIMO) and global optimization avoids coupling; PID requires the independent design of multiple controllers and is susceptible to interference between variables. MPC directly embeds constraints (such as the safe range of current and the safe range of speed) in the optimization to avoid the risk of exceeding the limit; PID requires external logic to process constraints, which can easily lead to integral saturation. MPC predicts future states through dynamic models and adjusts control actions in advance, significantly improving the response of lagging systems; PID relies only on current / historical error feedback, limiting its performance in lagging scenarios. MPC can handle nonlinear and time-varying models and supports online updates (such as real-time adjustments to position and image effects); PID relies on linear assumptions, has fixed parameters, and requires frequent parameter tuning in complex scenarios. MPC can comprehensively optimize multiple objectives such as tracking accuracy, energy consumption, and economic cost; PID only pursues the minimization of error.

[0086] MPC combines prediction and state estimation (such as designing Kalman filters) to actively compensate for measurable disturbances; PID relies on integral action and has a slower response under sudden disturbances.

Claims

1. A high-precision dynamic calibration algorithm for industrial CT, characterized by: Includes the following steps: Step 1: Based on the dynamic characteristics and kinematic constraints of the industrial CT motion control system, construct a state-space model that includes kinematic and dynamic factors. The state-space model includes the dynamic equations of the mechanical system, the state equations of the electrical system, and the state equations of the control system. Step 2: The model predictive control algorithm is adopted to predict the state sequence and input sequence of the system in the prediction time domain based on the discretized state space model. The calibration parameters are fitted by polynomial through a high-order fitting algorithm, and the motion platform is corrected in real time by combining the dynamic compensation algorithm. Step 3: Use high-precision sensors to collect command position data and actual position data of the CT control system in real time. After filtering and noise reduction preprocessing, the data is used for model verification and parameter optimization. Step 4: Compare the experimental data with the model prediction results, adjust the parameters of the high-order fitting algorithm to approximate the real dynamic characteristics of the system, and verify the calibration accuracy through third-party sensors and CT scan results. In step one, the kinematic factors include mechanical tolerances, machining errors, assembly errors, and reducer precision errors, while the dynamic factors include mass, inertia tensor, friction, and external disturbances. In step one, the model uses the zero-order preservation method to model the mechanical system and the forward discrete Euler method to model the electrical system. In step two, the dynamic compensation algorithm includes designing an extended state observer to estimate disturbances and friction, using an IP controller to compensate for the end effect error of the linear motor, and modeling the friction curve using a Gaussian fitting method. In step two, the objective function of model predictive control dynamically adjusts the weight parameters of displacement, velocity, and current according to the scanning scenario. The displacement weight is increased in the spiral CT / bias scan scenario, the velocity weight is increased in the ordinary CT / DR scan scenario, and the current weight is increased in the low-heat scenario. In step three, the high-precision sensor includes a laser interferometer with a resolution of not less than 1 nm and a linear phase grating encoder with a grating pitch of not more than 8 μm and an accuracy level of ±2 μm.

2. The high-precision dynamic calibration algorithm for industrial CT as described in claim 1, characterized in that: In step four, the image reconstruction quality is evaluated by peak signal-to-noise ratio, and the weight parameters are optimized based on Simulink simulation.

3. The high-precision dynamic calibration algorithm for industrial CT as described in any one of claims 1-2, characterized in that: It is used to improve the positioning accuracy, response speed and positioning stability during dynamic scanning, and to meet the internal structure inspection needs of coal, rock, lithium battery, CPU, PCB board, composite material and metal material under dynamic loading conditions.

Images