Intelligent calibration system for high-precision full-automatic centering instrument
Through the intelligent calibration system of multi-source data acquisition, edge computing and model training modules and adaptive execution control modules, the adaptability and resource waste problems of traditional centering instrument calibration technology in complex environments are solved, and high-precision and real-time calibration effects are achieved.
Patent Information
- Application Number
- CN202510695530.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-05-27
- Publication Date
- 2025-09-05
AI Technical Summary
Traditional high-precision fully automatic centering device calibration technology is difficult to adapt to the multi-dimensional interference and dynamic characteristic changes in complex industrial scenarios. It cannot update the calibration model in real time, cannot effectively model the changes in low-frequency system characteristics, and the calibration experience of different devices cannot be communicated, resulting in waste of resources.
It adopts a multi-source data acquisition module, an edge computing and model training module, and an adaptive execution control module, performs multimodal data fusion and anomaly detection through graph neural networks and Transformer models, generates a dynamic calibration matrix, and realizes sub-micron position adjustment through adaptive execution control.
It has achieved improved adaptability to multi-dimensional interference and dynamic characteristic changes, improved the accuracy, efficiency and reliability of the calibration system, and realized multi-device data sharing and cloud-based collaborative optimization.
Smart Images
Figure CN120595579A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of optical assembly and detection, and more particularly to an intelligent calibration system for a high-precision fully automatic centering instrument. Background Art
[0002] High-precision, fully automatic centering instruments are core equipment in optical manufacturing, precision mechanical assembly, and other fields. Their calibration accuracy directly determines product performance and process reliability. Traditional centering instrument calibration technology primarily relies on single optical sensor data and static control models, making it difficult to adapt to the multi-dimensional interference and dynamic characteristic changes in complex industrial scenarios. The specific bottlenecks are as follows:
[0003] (1) Most of them ignore cross-modal interference such as changes in ambient temperature and humidity, and mechanical structure deformation;
[0004] (2) The application of traditional calibration models (such as fixed parameter PID control) cannot be updated in real time;
[0005] (3) It is impossible to effectively model the changes in low-frequency system characteristics (such as mechanical structure aging);
[0006] (4) In large-scale production scenarios, the calibration experience of different centering instruments cannot be communicated with each other, and multi-device data sharing and cloud-based collaborative optimization cannot be achieved, resulting in repeated debugging of calibration parameters for similar workpieces, wasting a lot of computing power and time resources.
[0007] Therefore, how to provide an intelligent calibration system for a high-precision fully automatic centering instrument is a problem that those skilled in the art urgently need to solve. Summary of the Invention
[0008] In view of this, the present invention provides an intelligent calibration system for a high-precision fully automatic centering instrument, which is used to solve the technical problems existing in the above-mentioned prior art.
[0009] In order to achieve the above object, the present invention provides the following technical solutions:
[0010] An intelligent calibration system for a high-precision fully automatic centering device, comprising:
[0011] Multi-source data acquisition module for real-time monitoring of equipment status;
[0012] The edge computing and model training module uses graph neural networks (GNNs) for multimodal data fusion and anomaly detection, and uses the Transformer model combined with a spatiotemporal attention mechanism to dynamically generate an initial calibration matrix.
[0013] Adaptive execution control module to achieve sub-micron position adjustment based on the initial calibration matrix.
[0014] Furthermore, the multi-source data acquisition module includes:
[0015] Optical sensor, integrating high-precision autocollimator and laser interferometer, used to detect optical axis offset and position deviation;
[0016] Environmental sensors, including temperature sensors, humidity sensors, and vibration sensors, are used to collect working environment parameters;
[0017] Mechanical state sensors, consisting of encoders and strain gauges, are used to monitor mechanical structure deformation and servo motor position.
[0018] Furthermore, the edge computing and model training module uses a graph neural network (GNN) for multimodal data fusion and anomaly detection, and uses a Transformer model combined with a spatiotemporal attention mechanism to dynamically generate a calibration matrix, including:
[0019] Multimodal feature extraction unit, performs feature extraction;
[0020] The graph neural network (GNN) fusion unit builds a sensor relationship graph based on physical correlation to obtain fusion features;
[0021] The spatiotemporal attention mechanism unit captures long-term dependencies in time series data through a preset Transformer model and predicts the trend of calibration parameter changes;
[0022] The dynamic calibration matrix generation unit generates the initial calibration matrix M0 based on the fusion features and prediction results:
[0023]
[0024] Where Δx n , Δy n , Δz n is the linear displacement adjustment along the X, Y, and Z axes of the Cartesian coordinate system in the nth step of calibration; θ xn ,θ yn ,θ zn is the rotation angle adjustment around the X, Y, and Z axes in the nth step of calibration; t n The execution timestamp of the n-th step calibration action; σ n Uncertainty measure for the calibration parameters at step n.
[0025] Furthermore, the adaptive execution control module implements sub-micron position adjustment based on the initial calibration matrix, including:
[0026] Matrix analysis unit, which converts the initial calibration matrix into control instructions for each actuator;
[0027] Fuzzy PID controller dynamically adjusts control parameters based on real-time error and control instructions;
[0028] Multi-axis linkage execution unit drives high-precision servo motors and piezoelectric ceramic actuators for multi-dimensional adjustment;
[0029] A closed-loop feedback system verifies the adjustment results in real time through optical sensors and iteratively updates the initial calibration matrix based on a hierarchical optimization mechanism.
[0030] Furthermore, the hierarchical optimization mechanism of the closed-loop feedback system includes inner-loop optimization and outer-loop optimization;
[0031] The inner loop optimization compensates for high-frequency disturbances by fine-tuning PID parameters within a single calibration cycle;
[0032] The outer loop optimization modifies the initial calibration matrix based on the accumulated error after the calibration period.
[0033] Furthermore, the formula for the inner loop optimization is:
[0034]
[0035] Where, PID parameter matrix K = [K p ,K i ,K d ]It is dynamically adjusted according to the real-time error e(t) and only affects the execution instructions within the current control cycle.
[0036] Furthermore, the formula for the outer loop optimization is:
[0037]
[0038] Where, e i is the residual error vector after the i-th adjustment; w i is the time series weighting coefficient; λ is the regularization coefficient; M k is the calibration matrix of the kth iteration; J(M k ) is the loss function; H is the Hessian matrix; α is the learning rate.
[0039] Furthermore, the closed-loop feedback system also includes an anomaly detection mechanism that triggers calibration matrix reconstruction when the following conditions are detected:
[0040] The residual error is greater than the first residual error exceeding limit threshold for more than T cycles;
[0041] The calibration matrix update amount is greater than a second residual error exceeding limit threshold;
[0042] The multimodal data fusion results show that the system state has undergone a sudden change.
[0043] Furthermore, after the calibration matrix is reconstructed, the GNN data fusion and Transformer prediction process is re-executed through the edge computing module to generate a new initial calibration matrix M′0.
[0044] It can be seen from the above technical solution that compared with the existing technology, the present invention discloses an intelligent calibration system for high-precision fully automatic centering instruments. Through the mutual cooperation of multi-source data acquisition modules, edge computing and model training modules, and adaptive execution control modules, the edge computing and model training modules take the dominant role in prediction, and the adaptive execution control module is adaptable to time-varying factors, thereby improving the robustness of the system and realizing a technological leap from "single-modal static calibration" to "multi-modal dynamic intelligent calibration", effectively solving the accuracy, efficiency and reliability bottlenecks of traditional solutions. BRIEF DESCRIPTION OF THE DRAWINGS
[0045] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are merely embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.
[0046] Figure 1 Schematic diagram of the system structure of the present invention. DETAILED DESCRIPTION
[0047] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.
[0048] See also Figure 1 The embodiment of the present invention discloses an intelligent calibration system for a high-precision fully automatic centering instrument, comprising:
[0049] Multi-source data acquisition module for real-time monitoring of equipment status;
[0050] The edge computing and model training module uses graph neural networks (GNNs) for multimodal data fusion and anomaly detection, and uses the Transformer model combined with a spatiotemporal attention mechanism to dynamically generate an initial calibration matrix.
[0051] Adaptive execution control module to achieve sub-micron position adjustment based on the initial calibration matrix.
[0052] In a specific embodiment, the multi-source data acquisition module includes:
[0053] Optical sensor: Integrates a high-precision autocollimator (accuracy ≤ 0.1 arc second) and a laser interferometer (displacement accuracy ≤ 0.05 μm) to monitor optical axis offset and position deviation in real time;
[0054] Environmental sensors: temperature (accuracy ±0.1°C), humidity (accuracy ±3%RH), and vibration (accuracy ±0.01g) sensors to compensate for the impact of environmental factors on calibration;
[0055] Mechanical state sensor: encoder (resolution ≤ 0.001°) and strain gauge (accuracy ≤ 1με) to monitor mechanical structure deformation and servo motor position.
[0056] In a specific embodiment, the edge computing and model training module uses a graph neural network (GNN) for multimodal data fusion and anomaly detection, and uses a Transformer model combined with a spatiotemporal attention mechanism to dynamically generate a calibration matrix, including:
[0057] Multimodal feature extraction unit, performs feature extraction;
[0058] Optical data: 1D CNN extracts light spot features;
[0059] Time series data: LSTM captures the time series characteristics of environmental and mechanical parameters;
[0060] The graph neural network (GNN) fusion unit builds a sensor relationship graph based on physical correlation, with nodes representing sensors and edges representing physical correlation, and realizes cross-modal data fusion through GCN;
[0061] The spatiotemporal attention mechanism unit captures long-term dependencies in time series data through a preset Transformer model and predicts the trend of calibration parameter changes;
[0062] The dynamic calibration matrix generation unit generates the initial calibration matrix M0 based on the fusion features and prediction results:
[0063]
[0064] Where Δx n , Δy n , Δz n is the linear displacement adjustment along the X, Y, and Z axes of the Cartesian coordinate system in the nth step of calibration; θ xn ,θ yn ,θ zn is the rotation angle adjustment around the X, Y, and Z axes in the nth step of calibration; t n The execution timestamp of the n-th step calibration action; σ n Uncertainty measure for the calibration parameters at step n.
[0065] In a specific embodiment, the adaptive execution control module implements sub-micron position adjustment based on the initial calibration matrix, including:
[0066] Matrix analysis unit, which converts the initial calibration matrix into control instructions for each actuator;
[0067] Fuzzy PID controller dynamically adjusts control parameters based on real-time error and control instructions;
[0068] Multi-axis linkage execution unit drives high-precision servo motors and piezoelectric ceramic actuators for multi-dimensional adjustment;
[0069] A closed-loop feedback system verifies the adjustment results in real time through optical sensors and iteratively updates the initial calibration matrix based on a hierarchical optimization mechanism.
[0070] In one specific embodiment, the hierarchical optimization mechanism of the closed-loop feedback system includes inner-loop optimization and outer-loop optimization;
[0071] The inner loop optimization compensates for high-frequency disturbances by fine-tuning PID parameters within a single calibration cycle;
[0072] The outer loop optimization modifies the initial calibration matrix based on the accumulated error after the calibration period.
[0073] In a specific embodiment, the present invention adopts a two-layer optimization architecture to solve the iterative update problem of the calibration matrix:
[0074] Specifically, inner loop optimization (PID parameter fine-tuning) time scale: single control cycle (millisecond level) action object: PID controller parameter K p ,K i ,K d , the formula for inner loop optimization is:
[0075]
[0076] Where, PID parameter matrix K = [K p ,K i ,K d ]It is dynamically adjusted according to the real-time error e(t) and only affects the execution instructions within the current control cycle.
[0077] Specifically, outer loop optimization (calibration matrix correction):
[0078] Time scale: multiple control cycles (seconds to minutes);
[0079] Target: Initial calibration matrix;
[0080] Trigger condition: After a single calibration is completed (residual error converges);
[0081] Changes in system characteristics (such as accumulated temperature drift) are detected.
[0082] The formula for outer loop optimization is:
[0083]
[0084] Where, e i is the residual error vector after the i-th adjustment; w i is the time series weighting coefficient; λ is the regularization coefficient; M k is the calibration matrix of the kth iteration; J(M k ) is the loss function; H is the Hessian matrix; α is the learning rate.
[0085] In a specific embodiment, the closed-loop feedback system further includes an anomaly detection mechanism that triggers calibration matrix reconstruction when the following conditions are detected:
[0086] The residual error is greater than the first residual error exceeding limit threshold for more than T cycles;
[0087] The calibration matrix update amount is greater than a second residual error exceeding limit threshold;
[0088] The multimodal data fusion results show that the system state has undergone a sudden change.
[0089] In a specific embodiment, after the calibration matrix is reconstructed, the GNN data fusion and Transformer prediction process is re-executed through the edge computing module to generate a new initial calibration matrix M′0, rather than being corrected based solely on the historical matrix.
[0090] Fusion of the old and new calibration matrices via a Bayesian model:
[0091]
[0092] Among them, p(D|M) is the posterior probability of data D under model M.
[0093] Specifically, decoupling is achieved through hierarchical optimization, including: inner-loop optimization (PID parameter adjustment) to handle high-frequency dynamic disturbances and maintain system stability; outer-loop optimization (calibration matrix update) to handle low-frequency system characteristic changes and improve long-term accuracy.
[0094] Specifically, through the iteration of the calibration matrix, the initial calibration matrix \(M_0\) is generated by the edge computing module based on historical data; the closed-loop feedback system gradually corrects \(M_0\) through outer-loop optimization rather than completely regenerating it to maintain continuity.
[0095] Specifically, through the exception handling mechanism, when an anomaly beyond the correction range is detected, the edge computing module is triggered to re-predict, forming a complete closed loop.
[0096] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. Reference can be made to the common and similar parts between the various embodiments. For the devices disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple, and the relevant parts can be referred to the method description.
[0097] The above description of the disclosed embodiments is intended to enable one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be readily apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not limited to the embodiments shown herein but is intended to conform to the widest scope consistent with the principles and novel features disclosed herein.
Claims
1. An intelligent calibration system for a high-precision fully automatic centering instrument, characterized in that: include: Multi-source data acquisition module for real-time monitoring of equipment status; The edge computing and model training module uses graph neural networks (GNNs) for multimodal data fusion and anomaly detection, and uses the Transformer model combined with a spatiotemporal attention mechanism to dynamically generate an initial calibration matrix. Adaptive execution control module to achieve sub-micron position adjustment based on the initial calibration matrix.
2. The intelligent calibration system for a high-precision fully automatic centering instrument according to claim 1, characterized in that: The multi-source data acquisition module includes: Optical sensor, integrating high-precision autocollimator and laser interferometer, used to detect optical axis offset and position deviation; Environmental sensors, including temperature sensors, humidity sensors, and vibration sensors, are used to collect working environment parameters; Mechanical state sensors, consisting of encoders and strain gauges, are used to monitor mechanical structure deformation and servo motor position.
3. The intelligent calibration system for a high-precision fully automatic centering instrument according to claim 1, characterized in that: The edge computing and model training module uses a graph neural network (GNN) for multimodal data fusion and anomaly detection, and uses a Transformer model combined with a spatiotemporal attention mechanism to dynamically generate a calibration matrix, including: Multimodal feature extraction unit, performs feature extraction; The graph neural network (GNN) fusion unit builds a sensor relationship graph based on physical correlation to obtain fusion features; The spatiotemporal attention mechanism unit captures long-term dependencies in time series data through a preset Transformer model and predicts the trend of calibration parameter changes; The dynamic calibration matrix generation unit generates the initial calibration matrix M0 based on the fusion features and prediction results: Where Δx n , Δy n , Δz n is the linear displacement adjustment along the X, Y, and Z axes of the Cartesian coordinate system in the nth step of calibration; θ xn ,θ yn ,θ zn is the rotation angle adjustment around the X, Y, and Z axes in the nth step of calibration; t n The execution timestamp of the n-th step calibration action; σ n Uncertainty measure for the calibration parameters at step n.
4. The intelligent calibration system for a high-precision fully automatic centering device according to claim 1, characterized in that: The adaptive execution control module implements sub-micron position adjustment based on the initial calibration matrix, including: Matrix analysis unit, which converts the initial calibration matrix into control instructions for each actuator; Fuzzy PID controller dynamically adjusts control parameters based on real-time error and control instructions; Multi-axis linkage execution unit drives high-precision servo motors and piezoelectric ceramic actuators for multi-dimensional adjustment; A closed-loop feedback system verifies the adjustment results in real time through optical sensors and iteratively updates the initial calibration matrix based on a hierarchical optimization mechanism.
5. The intelligent calibration system for a high-precision fully automatic centering instrument according to claim 4, characterized in that: The hierarchical optimization mechanism of the closed-loop feedback system includes inner-loop optimization and outer-loop optimization; The inner loop optimization compensates for high-frequency disturbances by fine-tuning PID parameters within a single calibration cycle; The outer loop optimization modifies the initial calibration matrix based on the accumulated error after the calibration period.
6. The intelligent calibration system for a high-precision fully automatic centering device according to claim 5, characterized in that: The formula for the inner loop optimization is: Where, PID parameter matrix K = [K p ,K i ,K d ]It is dynamically adjusted according to the real-time error e(t) and only affects the execution instructions within the current control cycle.
7. The intelligent calibration system for a high-precision fully automatic centering device according to claim 5, characterized in that: The formula for the outer loop optimization is: M k+1 =M k +α·H -1 ·▽J(M k ) Where, e i is the residual error vector after the i-th adjustment; w i is the time series weighting coefficient; λ is the regularization coefficient; M k is the calibration matrix of the kth iteration; J(M k ) is the loss function; H is the Hessian matrix; α is the learning rate.
8. The intelligent calibration system for a high-precision fully automatic centering device according to claim 4, characterized in that: The closed-loop feedback system also includes an anomaly detection mechanism that triggers calibration matrix reconstruction when the following conditions are detected: The residual error is greater than the first residual error exceeding limit threshold for more than T cycles; The calibration matrix update amount is greater than a second residual error exceeding limit threshold; The multimodal data fusion results show that the system state has undergone a sudden change.
9. The intelligent calibration system for a high-precision fully automatic centering instrument according to claim 8, characterized in that: After the calibration matrix is reconstructed, the GNN data fusion and Transformer prediction process is re-executed through the edge computing module to generate a new initial calibration matrix M′0.