Intelligent compensation control method for precise feeding of linear guide rail numerical control grinding machine

Abstract

The invention provides an intelligent compensation control method for precise feeding of a linear guide rail numerical control grinding machine, which belongs to the technical field of precise machining of numerical control machine tools and comprises the following steps of: 1, scanning the surface of a linear guide rail to obtain a multi-dimensional data set containing temperature field data and morphology data; 2, feature extraction and analysis are conducted on the multi-dimensional data set through a deep learning model, a three-dimensional morphology graph of the surface of the linear guide rail is constructed, and the three-dimensional morphology graph is analyzed to generate a grinding path; 3, initial feeding compensation parameters are determined based on the grinding path, the three-dimensional shape graph and preset grinding parameters; 4, monitoring the temperature change and the surface appearance in the grinding process in real time, and adjusting an initial feeding compensation parameter; and 5, establishing a grinding quality evaluation model. The precision compensation problem caused by factors such as thermal deformation in linear rail grinding machining is solved, and the machining precision and the surface quality of the linear rail are remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of precision machining technology for CNC machine tools, and more particularly to a method for intelligent compensation control of precision feed for linear guide CNC grinding machines. Background Technology

[0002] In precision machinery manufacturing, linear guides are core components that ensure motion accuracy. Their straightness, parallelism, and surface quality directly determine the motion accuracy and stability of the entire machine. Grinding of linear guides is a key process to ensure their final accuracy. However, the intense friction between the grinding wheel and the workpiece during grinding generates a large amount of grinding heat, causing thermal deformation of the linear guides and seriously affecting the machining accuracy.

[0003] Traditional CNC grinding machines mainly rely on preset fixed machining parameters or operator experience for compensation, which makes it difficult to cope with complex grinding thermal deformation in real time and accurately, resulting in poor product dimensional consistency and high scrap rate. Most existing solutions focus on offline error compensation models and lack online perception and closed-loop feedback control of the temperature field and actual morphology of the workpiece during the grinding process, which cannot effectively suppress real-time dynamic thermal errors.

[0004] Therefore, the present invention provides a method for intelligent compensation control of precision feed in linear guide CNC grinding machines. Summary of the Invention

[0005] This invention provides an intelligent compensation control method for precision feed in linear guide CNC grinding machines, aiming to solve the problems of difficulty in precision compensation and unstable processing quality caused by factors such as thermal deformation in existing linear guide grinding processes. This invention combines infrared thermal imaging and three-dimensional topography scanning technology to perceive the thermal-shape coupling state of the grinding area in real time, and uses a deep learning model to intelligently analyze and generate the optimal compensation strategy to achieve closed-loop feedback control of precision feed, thereby improving the processing accuracy and surface quality of the linear guide, and increasing production efficiency and automation level.

[0006] This invention provides a method for intelligent compensation control of precision feed in a linear guide CNC grinding machine, comprising: Step 1: Scan the surface of the linear guide to obtain a multidimensional dataset containing temperature field data and morphology data; Step 2: Use a deep learning model to extract and analyze features from the multidimensional dataset, construct a three-dimensional topography map of the linear guide surface, and analyze the three-dimensional topography map to generate the grinding path; Step 3: Determine the initial feed compensation parameters based on the grinding path, 3D topography, and preset grinding parameters; Step 4: Monitor the temperature changes and surface morphology during the grinding process in real time, determine the real-time grinding status, and adjust the initial feed compensation parameters; Step 5: Establish a grinding quality evaluation model and use real-time collected surface feature data after grinding for closed-loop feedback control until the preset surface quality index is achieved.

[0007] Preferably, the surface of the linear guide is scanned to obtain a multidimensional dataset containing temperature field data and morphological data, including: The surface of the linear track is scanned using a pre-set acquisition device to obtain temperature field data. At the same time, three-dimensional topographic data is acquired using a pre-set sensor. Temperature field data and three-dimensional topography data are aligned using a spatiotemporal registration algorithm to establish a multidimensional dataset containing both temperature field data and topography data.

[0008] Preferably, a three-dimensional topographic map of the linear guide surface is constructed by using a deep learning model to extract and analyze features from the multidimensional dataset, and the grinding path is generated by analyzing the three-dimensional topographic map, including: A dual-channel deep neural network is constructed to extract features from a multidimensional dataset. The first channel network processes temperature field data and extracts thermodynamic features, while the second channel network processes three-dimensional topography data and extracts geometric features. Thermodynamic feature analysis is used to identify heat-affected regions, and geometric features are combined to determine heat-sensitive regions and geometric defect regions. The multi-objective cost function is determined based on temperature field data, three-dimensional topography data, and preset process constraints. The optimal path is determined by using a multi-objective cost function and a preset algorithm, combined with thermally sensitive areas and geometrically defective areas, and the grinding path is determined based on the optimal path.

[0009] Preferably, the initial feed compensation parameters are determined based on the grinding path, the three-dimensional topography image, and preset grinding parameters. The preset grinding parameters include preset grinding force and preset grinding speed, including: Determine several discrete points on the grinding path based on the grinding path; Based on the geometric features determined by the 3D topography map, the corresponding feature grinding parameters are determined in the preset feature-parameter database. The feature grinding parameters include: basic grinding force and basic grinding speed. The state parameters of the linear guide to be processed are obtained. Based on the preset grinding parameters, characteristic grinding parameters, and the state parameters of the linear guide to be processed, the initial compensation coefficients of each discrete point on the grinding path are determined, following the multi-factor coupled calculation logic: First, obtain the basic grinding force and basic grinding speed of the current discrete point. The product of the two is used as the grinding intensity of the current discrete point to reflect the instantaneous processing load of the current discrete point. Secondly, the wear influence factor is determined based on the friction coefficient between the grinding wheel and the linear guide, the average grinding time of all discrete points, and the preset grinding wheel hardness. The wear influence factor is then corrected with the current grinding intensity of the discrete point to obtain an intermediate value. Next, a hardness difference adjustment factor is determined based on the ratio of the average hardness of the rail to the local hardness of the current discrete point rail, which is used to correct the above intermediate value to obtain the corrected value. Meanwhile, based on the difference between the grinding temperature and the ambient temperature generated at the current discrete point, combined with the preset heat transfer coefficient, the contact area between the grinding wheel and the linear guide, and the preset heat loss coefficient, the heat effect correction factor is determined and coupled to the above correction value. Finally, the corrected value is compared and normalized with the preset reference value determined by the preset grinding force and the preset grinding speed to obtain the initial compensation coefficient of the current discrete point. The preset reference value is determined by the product of a preset grinding force and a preset grinding speed, and is used to characterize the theoretical processing strength under standard process conditions.

[0010] Preferably, real-time monitoring of temperature changes and surface morphology during the grinding process is used to determine the real-time grinding state and adjust the initial feed compensation parameters, including: Real-time temperature field data of the grinding area is acquired by an infrared thermal imager, and real-time surface morphology data of the grinding is simultaneously acquired by a laser displacement sensor and a structured light scanner. At the same time, the real-time temperature field data and the real-time surface morphology data of the grinding are preprocessed. Thermodynamic features are extracted from the temperature field data of real-time grinding, and geometric features are extracted from the morphology data of real-time grinding. By integrating thermodynamic and geometric features, a real-time processing state description vector is generated. The real-time machining status description vector is analyzed, and the initial feed compensation parameters are adjusted based on the analysis results.

[0011] Preferably, the real-time machining state description vector is analyzed, and the initial feed compensation parameters are adjusted based on the analysis results, including: Anomaly vectors are determined based on a preset threshold detection algorithm to identify real-time processing status description vectors. Based on a pre-defined vector-range database, the controllable range corresponding to the abnormal vector of each real-time processing state description vector is determined. If the abnormal vector of the real-time machining status description vector exceeds the controllable range, the grinding path segment where the abnormal vector is located will be marked as an abnormal path. Based on the anomaly vector, the corresponding preset adjustment strategy is called from the preset adjustment strategy database to adjust the initial feed compensation parameters of the anomaly path.

[0012] Preferably, a grinding quality evaluation model is established, and closed-loop feedback control is performed using real-time collected surface feature data after grinding until the preset surface quality indicators are achieved, including: Collect surface temperature field distribution data and surface geometric feature data after grinding, and simultaneously record process parameter data during the processing; The residual stress distribution data of the surface is determined based on the surface temperature field distribution data after grinding, the surface roughness parameters are extracted based on the surface geometric feature data after grinding, and a quality feature correlation matrix is ​​established in combination with the preset process parameters. Input the quality feature correlation matrix into the pre-trained grinding quality evaluation model, output the quality score distribution map, and determine the comprehensive quality score based on the quality score distribution map; Determine whether the overall quality score meets the quality requirements based on preset standards; For areas in the quality score distribution map that do not meet the quality requirements, a compensation processing path is generated, the process parameter data is adjusted, and an execution command is issued until the preset surface quality index is achieved.

[0013] Preferably, generating compensatory processing paths for areas in the quality score distribution map that do not meet quality requirements includes: Low-scoring regions were identified using a region growing algorithm based on the quality score distribution map. A spatial location mapping of the defect area is established based on the low-scoring area, and a compensation processing path is determined based on the morphological characteristics of the defect area and the preset feature-path database.

[0014] Compared with the prior art, the beneficial effects of this application are as follows: By combining infrared thermal imaging and deep learning technologies, adaptive control in the linear guide grinding process was achieved. Compared with existing technologies, it can monitor temperature changes and surface morphology in real time and precisely adjust grinding parameters, thereby ensuring the stability and consistency of surface quality. The closed-loop feedback control model continuously optimizes the grinding process, significantly improving grinding efficiency and quality, avoiding manual intervention and errors in traditional methods, and exhibiting strong adaptability and a high degree of automation. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0016] Figure 1 This is a flowchart illustrating the intelligent compensation control method for precision feed of a linear guide CNC grinding machine provided in an embodiment of the present invention. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0018] Example 1

[0019] This invention provides a method for intelligent compensation control of precision feed in linear guide CNC grinding machines, such as... Figure 1 As shown, it includes: Step 1: Scan the surface of the linear guide to obtain a multidimensional dataset containing temperature field data and morphology data; Step 2: Use a deep learning model to extract and analyze features from the multidimensional dataset, construct a three-dimensional topography map of the linear guide surface, and analyze the three-dimensional topography map to generate the grinding path; Step 3: Determine the initial feed compensation parameters based on the grinding path, 3D topography, and preset grinding parameters; Step 4: Monitor the temperature changes and surface morphology during the grinding process in real time, determine the real-time grinding status, and adjust the initial feed compensation parameters; Step 5: Establish a grinding quality evaluation model and use real-time collected surface feature data after grinding for closed-loop feedback control until the preset surface quality index is achieved.

[0020] In this embodiment, temperature field data refers to the temperature distribution information of the linear guide surface obtained through infrared thermal imaging technology. This data reflects the heat distribution generated on the linear guide surface during grinding due to friction and pressure changes. It enables real-time monitoring of temperature changes, thereby helping to determine the grinding effect and surface quality. For example, during grinding, certain areas of the linear guide surface may generate higher temperatures due to excessive friction. This information will be displayed as data showing temperature values ​​at different locations. For instance, the temperature value in one area might be 35°C, while in another area it might be 28°C, reflecting the heat distribution. In this embodiment, morphological data refers to information about the surface morphology of the linear guide obtained through scanning technology, including surface roughness, unevenness, and wear areas. This data is typically obtained through laser scanning, structured light, or other methods and can generate three-dimensional morphological images to help analyze surface changes during the grinding process. For example, suppose that during grinding, a small portion of the linear guide surface is smoothed out, while other parts remain highly raised. Scanning can obtain three-dimensional data of this surface, showing the height differences between different areas. For instance, the scan results may show that the surface height of one area is 0.5 mm, while that of another area is 0.2 mm.

[0021] In this embodiment, the grinding path refers to the trajectory of the tool moving along the surface of the linear guide during the grinding process. After analyzing the topography data using a deep learning model, an optimal grinding path can be generated to ensure uniform grinding of the linear guide surface while avoiding quality problems caused by localized over-grinding or insufficient grinding. For example, in actual operation, the grinding path may be a curve or multiple parallel lines on the linear guide surface, indicating which routes the grinding tool should follow from the starting point to the ending point. Assuming there are raised sections on the linear guide, the grinding path will choose to follow the highest point of the raised section to ensure uniform surface grinding.

[0022] In this embodiment, the preset surface quality index refers to the surface quality standard expected to be achieved during the grinding process, including straightness, parallelism, surface roughness, etc., which is used to evaluate whether the linear guide after grinding meets the requirements of high-precision motion.

[0023] The beneficial effects of the above technical solution are as follows: By combining infrared thermal imaging and deep learning technology, adaptive control in the linear guide grinding process is achieved. Compared with existing technologies, it can monitor temperature changes and surface morphology in real time and accurately adjust grinding parameters, thereby ensuring the stability and consistency of surface quality. The closed-loop feedback control model continuously optimizes the grinding process, significantly improving grinding efficiency and quality, avoiding manual intervention and errors in traditional methods, and exhibiting strong adaptability and a high degree of automation.

[0024] Example 2

[0025] This invention provides a method for intelligent compensation control of precision feed in a linear guide CNC grinding machine. The method involves scanning the surface of the linear guide to obtain a multidimensional dataset containing temperature field data and morphology data, including: The surface of the linear track is scanned using a pre-set acquisition device to obtain temperature field data. At the same time, three-dimensional topographic data is acquired using a pre-set sensor. Temperature field data and three-dimensional topography data are aligned using a spatiotemporal registration algorithm to establish a multidimensional dataset containing both temperature field data and topography data.

[0026] In this embodiment, the preset acquisition device is an infrared thermal imager, such as a FLIR A655sc infrared thermal imager (sampling frequency 200Hz).

[0027] In this embodiment, temperature field data and topography data are aligned using a spatiotemporal registration algorithm, including: time synchronization: ensuring synchronized data acquisition from multiple sensors through hardware triggering; coordinate unification: establishing a spatial transformation relationship between temperature field and topography data based on a calibration board; feature matching: extracting common feature points (such as edges and marker points) for alignment optimization; and dynamic compensation: correcting time delay errors based on motion parameters to complete data fusion.

[0028] In this embodiment, precise control of the linear guide grinding process is achieved by combining infrared thermal imaging and multidimensional data collected by sensors with a spatiotemporal registration algorithm. Compared with existing technologies, the alignment of temperature field data and surface morphology data enables the system to more accurately monitor and adjust the surface quality during the grinding process. Through real-time acquisition of multidimensional data, the system can optimize the grinding path and provide real-time feedback adjustments to ensure that preset surface quality indicators are met, significantly improving the automation level and quality stability of the grinding process.

[0029] The beneficial effects of the above technical solution are as follows: By combining infrared thermal imaging and multidimensional data collected by sensors with a spatiotemporal registration algorithm, precise control of the linear guide grinding process is achieved. Compared with existing technologies, the alignment of temperature field data and surface morphology data enables the system to more accurately monitor and adjust the surface quality during the grinding process. Through real-time acquisition of multidimensional data, the system can optimize the grinding path and provide real-time feedback adjustments to ensure that preset surface quality indicators are met, significantly improving the automation level and quality stability of grinding.

[0030] Example 3: This invention provides a precision feed intelligent compensation control method for linear guide CNC grinding machines. It uses a deep learning model to extract and analyze features from a multidimensional dataset, constructs a three-dimensional topographic map of the linear guide surface, and analyzes the three-dimensional topographic map to generate a grinding path, including: A dual-channel deep neural network is constructed to extract features from a multidimensional dataset. The first channel network processes temperature field data and extracts thermodynamic features, while the second channel network processes three-dimensional topography data and extracts geometric features. Thermodynamic feature analysis is used to identify heat-affected regions, and geometric features are combined to determine heat-sensitive regions and geometric defect regions. The multi-objective cost function is determined based on temperature field data, three-dimensional topography data, and preset process constraints. The optimal path is determined by using a multi-objective cost function and a preset algorithm, combined with thermally sensitive areas and geometrically defective areas, and the grinding path is determined based on the optimal path.

[0031] In this embodiment, feature extraction includes: First channel: using a 3D convolutional network to process time-series temperature field data, extracting thermodynamic features such as temperature gradient and heat flow distribution through multi-layer convolution; Second channel: using a point cloud convolutional network to analyze three-dimensional morphology, capturing geometric features such as curvature and height difference through geometric convolutional layers.

[0032] In this embodiment, the thermally affected region (TIR) ​​is identified based on thermodynamic feature analysis, and the thermally sensitive region and geometric defect region are determined by combining morphological features. The steps include: TIR identification: Thermally sensitive region identification, based on infrared thermal imaging data, is achieved through temperature gradient analysis, and the TIR boundary is determined using an adaptive threshold segmentation method; Geometric defect region detection, based on three-dimensional morphological scanning data, defect regions are identified through curvature analysis and height deviation calculation, and regions are classified and prioritized according to defect feature parameters; Thermo-mechanical coupling region calibration, spatially superimposed analysis of the thermally sensitive region and the geometric defect region; Identification of key processing areas and marking of special process requirements; Dynamic verification and correction, verifying the accuracy of region division through real-time monitoring data, and dynamically updating the region division results based on new findings during processing.

[0033] In this embodiment, the preset process constraints comprehensively cover multiple aspects such as material property constraints, machining quality constraints, process capability constraints, and efficiency and economic constraints. Material property constraints aim to limit the maximum machining temperature through a heat-sensitive threshold, constrain the hardness ratio between the grinding wheel and the linear guide through a hardness matching range, and set a critical point for plastic deformation to prevent material deformation caused by excessive grinding. Machining quality constraints require that the upper limit of surface roughness reach a precision-grade finish, control the internal stress caused by machining within the residual stress threshold, and maintain the integrity of geometric features. Process capability constraints are limited by the machine tool's dynamic parameters (such as maximum acceleration / speed), the wear limit of the grinding wheel (such as effective working radius), and cooling conditions (such as maximum coolant flow rate). Efficiency and economic constraints limit the amount removed in a single pass, path overlap rate, and total machining time. These constraints together ensure the high dynamic response and long service life required for the linear guide as a precision transmission component. In this embodiment, the multi-objective cost function is a mathematical expression used to simultaneously optimize multiple conflicting objectives. It is primarily used to balance multiple key indicators in path planning, and the objectives include: Path length objective: Minimize the total grinding path length to improve efficiency; Thermal impact objective: Minimize the heat load on heat-sensitive areas and the defect handling objective: maximize coverage of high-priority defects.

[0034] In this embodiment, the grinding path is determined by combining the heat-sensitive area and the geometric defect area: the identified heat-sensitive area is avoided; the geometric defect area is processed first and set as a high priority. In this embodiment, the optimal path determination steps are as follows: Target modeling: Construct a multi-objective function that includes path length, thermal effects, and defect handling; Constraint setting: Define boundary conditions by combining material properties and process requirements; Algorithm solution: Use an improved A* algorithm to search for the Pareto optimal solution set; Decision optimization: Dynamically adjust weights based on real-time data and output the final path, i.e., the grinding path.

[0035] The beneficial effects of the above technical solution are as follows: By analyzing multidimensional datasets through a dual-channel deep neural network, thermodynamic and geometric features are extracted separately, enabling accurate identification of heat-affected zones and geometric defect areas. Compared with existing technologies, this invention constructs a multi-objective cost function by combining temperature field data, three-dimensional morphology data, and process constraints, which can intelligently optimize the grinding path, ensuring accuracy and efficiency during the grinding process. The system can adjust the grinding path in real time, improving surface quality consistency and processing accuracy, reducing manual intervention and errors, and enhancing the level of automation control.

[0036] Example 4: This invention provides a method for intelligent compensation control of precision feed in a linear guide CNC grinding machine, which determines initial feed compensation parameters based on the grinding path and preset grinding parameters, including: The initial feed compensation parameters are determined based on the grinding path, 3D topography, and preset grinding parameters, including: Determine several discrete points on the grinding path based on the grinding path; Based on the geometric features determined by the 3D topography map, the corresponding feature grinding parameters are determined in the preset feature-parameter database. The feature grinding parameters include: basic grinding force and basic grinding speed. Obtain the state parameters of the linear guide to be processed, and determine the initial compensation coefficient for each discrete point on the grinding path based on the preset grinding parameters, characteristic grinding parameters, and the state parameters of the linear guide to be processed:

[0037] in, Let be the initial compensation coefficient for the i-th discrete point. The preset grinding force, The preset grinding speed, The base grinding force at the i-th discrete point. Let be the base grinding speed at the i-th discrete point. The preset mold hardness. The coefficient of friction between the mold and the linear guide. The preset heat loss coefficient, The average hardness of the linear guide. Let be the local hardness of the line track at the i-th discrete point. The preset heat transfer coefficient, This represents the contact area between the mold and the linear guide. For ambient temperature, Let be the grinding temperature generated at the i-th discrete point. The average grinding time is the average grinding time for all discrete points.

[0038] In this embodiment, determining several discrete points based on the grinding path includes: Step 1: Path segmentation processing. Based on the curvature variation characteristics of the grinding path, the continuous path is divided into straight line segments and curved line segments; key control points are added at curvature abrupt changes (such as inflection points and connection points); Step 2: Adaptive discretization. Based on preset machining accuracy requirements, each path segment is discretized with equal error: straight line segments are sampled with a fixed step size; the sampling density of curved line segments is dynamically adjusted according to the radius of curvature (the larger the curvature, the smaller the point spacing); Step 3: Process parameter association. Three-dimensional topography data and temperature field data are associated for each discrete point; points requiring special process treatment (such as defect concentration areas and heat-sensitive areas) are marked; Step 4: Dynamic optimization verification. The rationality of the discrete point distribution is verified through kinematic simulation; ensuring that the parameter changes between adjacent points meet the machine tool dynamics constraints.

[0039] The preset grinding parameters include: preset grinding speed, preset grinding force, preset grinding wheel hardness, and average hardness of the linear guide. The state parameters of the grinding linear guide include: friction coefficient between the grinding wheel and the linear guide, contact area between the grinding wheel and the linear guide, local hardness of the linear guide, and average grinding time for all discrete points. The characteristic grinding parameters include: basic grinding force and basic grinding speed.

[0040] In this embodiment, the preset feature-parameter database is a mapping relationship library between morphological features and process parameters. For example: flat area (curvature ≤ 0.01 / mm): basic force 5N, speed 20mm / s; raised area (curvature 0.01-0.05 / mm): force 8N, speed 15mm / s; concave area (curvature ≥ 0.05 / mm): force 12N, speed 10mm / s.

[0041] In this embodiment, the friction coefficient between the mold and the linear guide is a dimensionless value, which can be obtained through experimental measurement or by consulting relevant material friction coefficient handbooks. Different combinations of mold and linear guide materials have different friction coefficient values. In this embodiment, the preset heat transfer coefficient is a parameter related to the grinding environment and the properties of the contact materials. The heat transfer coefficient can be obtained through experimental measurement or by consulting relevant heat transfer handbooks. For specific grinding environments (such as air medium, temperature, humidity, etc.) and the combination of contact materials between the grinding wheel and the linear guide, there are corresponding heat transfer coefficient values. In some cases, if an accurate value cannot be obtained directly, it can be estimated through approximate calculation or by referring to empirical values ​​for similar environments and material combinations.

[0042] In this embodiment, ambient temperature refers to the temperature of the surrounding environment where the grinding operation takes place.

[0043] In this embodiment, the preset coefficient-parameter database establishes a mapping relationship between grinding coefficients and process parameters through experimental data. For example, the coefficient range is 0.5 (low strength) to 1.5 (high strength). The parameter matching rule is: coefficient ≤ 0.8: reduce the force by 20% and increase the speed by 15% (to avoid heat accumulation); 0.8 < coefficient ≤ 1.2: maintain the baseline parameter; coefficient > 1.2: increase the force by 30% and reduce the speed by 20% (to deal with high hardness areas).

[0044] The beneficial effects of the above technical solution are as follows: By accurately calculating the initial compensation coefficient of each discrete point on the grinding path, and combining it with the three-dimensional topography map, feature-parameter database, and linear guide state parameters, the process of determining grinding parameters is optimized. Compared with existing technologies, the system can intelligently adjust the grinding force and speed according to factors such as linear guide surface characteristics and local hardness, improving the accuracy and consistency of the grinding process. By comprehensively considering multiple factors, thermal control and surface quality during the grinding process are ensured, significantly improving automation and adaptability, and reducing human operation errors and uneven grinding problems.

[0045] Example 5

[0046] This invention provides a method for intelligent compensation control of precision feed in a linear guide CNC grinding machine. This method monitors temperature changes and surface morphology during the grinding process in real time, determines the real-time grinding state, and adjusts the initial feed compensation parameters, including: Real-time temperature field data of the grinding area is acquired by an infrared thermal imager, and real-time surface morphology data of the grinding is simultaneously acquired by a laser displacement sensor and a structured light scanner. At the same time, the real-time temperature field data and the real-time surface morphology data of the grinding are preprocessed. Thermodynamic features are extracted from the temperature field data of real-time grinding, and geometric features are extracted from the morphology data of real-time grinding. By integrating thermodynamic and geometric features, a real-time processing state description vector is generated. The real-time machining status description vector is analyzed, and the initial feed compensation parameters are adjusted based on the analysis results.

[0047] In this embodiment, the real-time machining state description vector is a comprehensive feature vector representing the current state during the grinding process. It combines the thermodynamic and geometric features of temperature field data and surface morphology data. It provides real-time feedback on the grinding process, helping the system analyze the interrelationships between surface quality, temperature changes, and morphological features. Through this vector, the system can comprehensively understand the current grinding state and adjust grinding parameters based on this information, ensuring efficient and precise grinding. For example, suppose that during the grinding process, temperature field data (acquired via an infrared thermal imager) shows a high temperature in a certain area, reaching 40°C, while surface morphology data (acquired via a laser displacement sensor) shows a large depression on the surface of that area. By extracting these data, the system generates a real-time machining state description vector containing the following elements: thermodynamic features (such as temperature value, heat-affected zone) and geometric features (such as depression depth, surface roughness). This description vector reflects the current machining state in real time and is transmitted to the control system for analysis and adjustment of initial parameters such as grinding force and speed to achieve the preset surface quality standard.

[0048] In this embodiment, the preprocessing steps include: noise filtering: using Gaussian filtering to eliminate sensor noise; data alignment: unifying the coordinate system of temperature field and morphology data through spatiotemporal registration; missing data compensation: supplementing missing data in occluded areas based on interpolation algorithms; and normalization processing: scaling multi-source data to a uniform dimension.

[0049] The beneficial effects of the above technical solution are as follows: By synchronously acquiring temperature field data and surface morphology data, and extracting thermodynamic and geometric features, a real-time processing state description vector is generated, which can accurately reflect the current state during the grinding process. Compared with existing technologies, the system can adjust the initial feed compensation parameters in real time by analyzing this description vector, ensuring temperature control and surface quality optimization during the grinding process, improving the automation level of the grinding process, reducing errors, improving grinding accuracy and efficiency, and ensuring high-quality final surface treatment.

[0050] Example 6

[0051] This invention provides a method for intelligent compensation control of precision feed in a linear guide CNC grinding machine. It analyzes the real-time machining state description vector and adjusts the initial feed compensation parameters based on the analysis results, including: Anomaly vectors are determined based on a preset threshold detection algorithm to identify real-time processing status description vectors. Based on a pre-defined vector-range database, the controllable range corresponding to the abnormal vector of each real-time processing state description vector is determined. If the abnormal vector of the real-time machining status description vector exceeds the controllable range, the grinding path segment where the abnormal vector is located will be marked as an abnormal path. Based on the anomaly vector, the corresponding preset adjustment strategy is called from the preset adjustment strategy database to adjust the initial feed compensation parameters of the anomaly path.

[0052] In this embodiment, the controllable range refers to the state values ​​within the preset parameter variation range during the grinding process, representing the normal working area of ​​the grinding process. Specifically, it refers to the state range within which the abnormal vector of each real-time machining state description vector can still be controlled by adjusting the initial feed compensation parameters when the abnormal vector fluctuates within a certain range. If the abnormal vector exceeds this range, it indicates that the current grinding process is abnormal and requires intervention and adjustment. Suppose that during the grinding process, an abnormal temperature value and morphological feature in a certain real-time machining state description vector are detected as abnormal. Through a preset threshold detection algorithm, the system finds that the abnormal temperature value exceeds the controllable range. For example, the temperature exceeds 50°C (the set upper limit is 45°C). In this case, the system will identify that the state exceeds the preset controllable range and mark it as an abnormal path. Then, by adjusting the strategy database, the system will take preset adjustment measures, such as slowing down the grinding speed or adjusting the grinding force, to restore the normal machining state.

[0053] In this embodiment, the preset threshold detection algorithm is based on the material properties and process standards to pre-set warning thresholds for key parameters (such as temperature gradient, roughness change rate, etc.), and identifies abnormal data vectors that deviate from the normal process window through a sliding window statistical method.

[0054] In this embodiment, the preset vector-range database stores the allowable fluctuation range of parameters under different processing conditions (e.g., the upper limit of temperature fluctuation of titanium alloy in the fine grinding stage is ±8℃), and the controllable boundary value of the current abnormal vector is matched by matrix query.

[0055] In this embodiment, the preset adjustment strategy database contains correction strategies for typical abnormal operating conditions (such as the "speed reduction + intermittent cooling" strategy when there is local overheating). The abnormal type and adjustment action are associated through a decision tree model, which supports dynamic weight adjustment.

[0056] In this embodiment, the initial feed compensation parameters for abnormal paths are adjusted. For example, when the temperature exceeds the limit, the grinding speed is reduced or the cooling intensity is increased; when the morphology is not up to standard, the grinding force or path overlap rate is adjusted; the parameter adjustment amount is generated by the fuzzy inference system to ensure the smoothness of the adjustment. The beneficial effects of the above technical solution are as follows: By monitoring the processing status in real time and combining it with a preset threshold detection algorithm, the system can accurately identify abnormal vectors and determine the grinding path when they exceed the controllable range. Compared with existing technologies, the system automatically adjusts the grinding parameters of abnormal paths by calling the adjustment strategy database, ensuring the stability of the grinding process and surface quality, improving adaptive adjustment capabilities, reducing manual intervention and errors, improving the accuracy and efficiency of the processing process, and guaranteeing high-quality surface treatment of the linear guide.

[0057] Example 7

[0058] This invention provides a method for intelligent compensation control of precision feed in a linear guide CNC grinding machine. It establishes a grinding quality evaluation model and uses real-time collected surface feature data after grinding for closed-loop feedback control until a preset surface quality index is achieved. The method includes: Collect surface temperature field distribution data and surface geometric feature data after grinding, and simultaneously record process parameter data during the processing; The residual stress distribution data of the surface is determined based on the surface temperature field distribution data after grinding, the surface roughness parameters are extracted based on the surface geometric feature data after grinding, and a quality feature correlation matrix is ​​established in combination with the preset process parameters. Input the quality feature correlation matrix into the pre-trained grinding quality evaluation model, output the quality score distribution map, and determine the comprehensive quality score based on the quality score distribution map; Determine whether the overall quality score meets the quality requirements based on preset standards; For areas in the quality score distribution map that do not meet the quality requirements, a compensation processing path is generated, the process parameter data is adjusted, and an execution command is issued until the preset surface quality index is achieved.

[0059] In this embodiment, the preset standard is a preset comprehensive quality score threshold; In this embodiment, a quality feature correlation matrix is ​​established, including: morphology analysis: extracting surface roughness (Sa / Sz) parameters through wavelet transform; process correlation: matching roughness data with process parameters (force, speed) at corresponding points; matrix construction: establishing a multidimensional correlation matrix between process parameters and roughness; and pattern mining: identifying key influencing factors through principal component analysis (PCA).

[0060] In this embodiment, the preset process parameters refer to a set of baseline process parameters pre-set according to material properties, processing objectives, and equipment capabilities, used to guide the generation and execution of compensated processing paths. Its core parameters include, but are not limited to, the following: basic processing parameters: grinding force, grinding speed, path overlap rate (dynamically adjusted based on defect area, e.g., 20%-80%), thermal management parameters: cooling intensity (dynamically adjusted based on real-time temperature field data), intermittent time: pause interval set to prevent local overheating, quality control parameters, maximum allowable residual stress, surface roughness threshold (e.g., guide rail sliding surface Ra≤0.1μm, side surface Ra≤0.4μm), straightness tolerance ≤0.005mm / m, adjustment logic, matching optimized parameter combinations from the preset parameter library based on the defect type (e.g., pits, protrusions) and severity of non-compliant areas in the quality scoring distribution map; and avoiding parameter abrupt changes through a gradient adjustment strategy (e.g., gradually increasing the force by 10%-20%).

[0061] In this embodiment, the trained grinding quality evaluation model is a multi-input deep neural network, with the following structure: Input layer: receives the quality feature correlation matrix, including: residual stress distribution (calculated by inversion from temperature field data), surface roughness parameters (Sa / Sz, etc., extracted from morphology data), and process parameters (time-series data such as grinding force and speed); Feature fusion layer: extracts spatial features through a convolution module and extracts time-series features through an LSTM module; Output layer: Branch 1: outputs a pixel-level quality score distribution map (resolution consistent with the line track scanning data); Branch 2: outputs the overall comprehensive quality score (0-1 standardized value). 2. Model application method: Online inference: real-time collected data is preprocessed and then input into the model, forward computation time <50ms; Dynamic update: after every 10 processing cycles, the model parameters are fine-tuned with new data. 3. Output definition: Quality score distribution map: each pixel corresponds to a quality score of 0-100 points, scoring criteria: ≥90 points: Standard area (green mark), 70-89 points: Acceptable area (yellow mark). The comprehensive quality score Q is determined based on weighted summation. Judgment standard: Q_total≥85 points is considered as overall qualified. 4. Scoring determination logic, spatial weight allocation: the weight of key functional areas (such as linear guide sliding contact surface) is increased by 30%, defect aggregation analysis: clustering of continuous non-standard areas and deducting points according to the defect area, process correlation correction: when process parameters exceed the recommended range, a penalty coefficient of 10-20% is applied to the score.

[0062] The beneficial effects of the above technical solution are as follows: By establishing a grinding quality evaluation model and combining temperature field, geometric features, and process parameter data, closed-loop feedback control is achieved. Compared with existing technologies, the system can evaluate grinding quality in real time, accurately identify substandard areas based on the quality score distribution map, and generate compensation machining paths for adjustment, thereby improving the control accuracy of surface quality, reducing manual intervention, and improving the automation level and stability of the grinding process through continuous optimization of process parameters, ensuring high-quality surface treatment of the linear guide.

[0063] Example 8

[0064] This invention provides a method for intelligent compensation control of precision feed in a linear guide CNC grinding machine, which generates compensation machining paths for areas in a quality score distribution map where the score pairs do not meet quality requirements, including: Low-scoring regions were identified using a region growing algorithm based on the quality score distribution map. A spatial location mapping of the defect area is established based on the low-scoring area, and a compensation processing path is determined based on the morphological characteristics of the defect area and the preset feature-path database.

[0065] In this embodiment, the spatial location mapping steps for the defect area are as follows: Coordinate transformation: Aligning the quality score distribution map with the coordinate system of the three-dimensional topography data; Area marking: Dividing the boundaries of low-scoring areas using a clustering algorithm; Feature association: Extracting the geometric features (depth / curvature) of the defect area and associating them with spatial coordinates; Mapping output: Generating a three-dimensional distribution map of the defect area with location encoding.

[0066] In this embodiment, the region growing algorithm is an image segmentation algorithm that starts from a seed point and gradually merges adjacent pixels according to a similarity criterion to form a continuous region. In this invention, it is used to aggregate defective regions with similar quality scores. Specific implementation steps (example): Seed point selection: Select the point with the lowest score (e.g., score < 70) from the quality score distribution map as the initial seed. Growth condition: the score difference between adjacent pixels ≤ 10 points (to ensure consistency of defect areas); Spatial continuity constraint (merging only adjacent pixels). Termination condition: Region area ≥ 5mm² (to avoid noise interference); growth stops when encountering a boundary with a score ≥ 80. Output result: Mark all consecutive low-scoring areas as defect areas to be compensated (e.g., Figure 1 (Middle red block).

[0067] In this embodiment, the preset feature-path database structure includes: shape feature type, feature quantization value, compensation path strategy, and applicable scenario. For example, shape feature type: shallow pit (depth ≤ 0.1 mm), feature quantization value: curvature ≤ 0.03 / mm, compensation path strategy: spiral path, 50% overlap rate, and applicable scenario: minor defect repair.

[0068] The beneficial effects of the above technical solution are as follows: By accurately locating low-scoring areas based on the quality score distribution map and region growing algorithm, and generating compensation processing paths by combining the morphological characteristics of defective areas, the substandard areas are effectively repaired. Compared with existing technologies, this optimizes the defect detection and compensation path generation process, improves compensation accuracy, reduces unnecessary processing, and improves grinding quality. This technology effectively reduces manual intervention and repair costs, and enhances the consistency and quality control capabilities of linear guide surface treatment.

[0069] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for intelligent compensation control of precision feed in linear guide CNC grinding machines, characterized in that, include: Step 1: Scan the surface of the linear guide to obtain a multidimensional dataset containing temperature field data and morphology data; Step 2: Use a deep learning model to extract and analyze features from the multidimensional dataset, construct a three-dimensional topography map of the linear guide surface, and analyze the three-dimensional topography map to generate the grinding path; Step 3: Determine the initial feed compensation parameters based on the grinding path, 3D topography, and preset grinding parameters; Step 4: Monitor the temperature changes and surface morphology during the grinding process in real time, determine the real-time grinding status, and adjust the initial feed compensation parameters; Step 5: Establish a grinding quality evaluation model and use real-time collected surface feature data after grinding for closed-loop feedback control until the preset surface quality index is achieved.

2. The intelligent compensation control method for precision feed of a linear guide CNC grinding machine according to claim 1, characterized in that, The surface of the linear guide is scanned to obtain a multidimensional dataset containing temperature field data and morphological data, including: The surface of the linear track is scanned using a pre-set acquisition device to obtain temperature field data. At the same time, three-dimensional topographic data is acquired using a pre-set sensor. Temperature field data and three-dimensional topography data are aligned using a spatiotemporal registration algorithm to establish a multidimensional dataset containing both temperature field data and topography data.

3. The intelligent compensation control method for precision feed of a linear guide CNC grinding machine according to claim 2, characterized in that, Feature extraction and analysis of a multidimensional dataset are performed using a deep learning model to construct a 3D topographic map of the linear guide surface. The 3D topographic map is then analyzed to generate a grinding path, including: A dual-channel deep neural network is constructed to extract features from a multidimensional dataset. The first channel network processes temperature field data and extracts thermodynamic features, while the second channel network processes three-dimensional topography data and extracts geometric features. Thermodynamic feature analysis is used to identify heat-affected regions, and geometric features are combined to determine heat-sensitive regions and geometric defect regions. The multi-objective cost function is determined based on temperature field data, three-dimensional topography data, and preset process constraints. The optimal path is determined by using a multi-objective cost function and a preset algorithm, combined with thermally sensitive areas and geometrically defective areas, and the grinding path is determined based on the optimal path.

4. The intelligent compensation control method for precision feed of a linear guide CNC grinding machine according to claim 3, characterized in that, The initial feed compensation parameters are determined based on the grinding path, 3D topography, and preset grinding parameters. These preset grinding parameters include preset grinding force and preset grinding speed, and include: Determine several discrete points on the grinding path based on the grinding path; Based on the geometric features determined by the 3D topography map, the corresponding feature grinding parameters are determined in the preset feature-parameter database. The feature grinding parameters include: basic grinding force and basic grinding speed. The state parameters of the linear guide to be processed are obtained. Based on the preset grinding parameters, characteristic grinding parameters, and the state parameters of the linear guide to be processed, the initial compensation coefficients of each discrete point on the grinding path are determined, following the multi-factor coupled calculation logic: First, obtain the basic grinding force and basic grinding speed of the current discrete point. The product of the two is used as the grinding intensity of the current discrete point to reflect the instantaneous processing load of the current discrete point. Secondly, the wear influence factor is determined based on the friction coefficient between the grinding wheel and the linear guide, the average grinding time of all discrete points, and the preset grinding wheel hardness. The wear influence factor is then corrected with the current grinding intensity of the discrete point to obtain an intermediate value. Next, a hardness difference adjustment factor is determined based on the ratio of the average hardness of the rail to the local hardness of the current discrete point rail, which is used to correct the above intermediate value to obtain the corrected value. Meanwhile, based on the difference between the grinding temperature and the ambient temperature generated at the current discrete point, combined with the preset heat transfer coefficient, the contact area between the grinding wheel and the linear guide, and the preset heat loss coefficient, the heat effect correction factor is determined and coupled to the above correction value. Finally, the corrected value is compared and normalized with the preset reference value determined by the preset grinding force and the preset grinding speed to obtain the initial compensation coefficient of the current discrete point. The preset reference value is determined by the product of a preset grinding force and a preset grinding speed, and is used to characterize the theoretical processing strength under standard process conditions.

5. The intelligent compensation control method for precision feed of a linear guide CNC grinding machine according to claim 1, characterized in that, Real-time monitoring of temperature changes and surface morphology during the grinding process to determine the real-time grinding state and adjust the initial feed compensation parameters, including: Real-time temperature field data of the grinding area is acquired by an infrared thermal imager, and real-time surface morphology data of the grinding is simultaneously acquired by a laser displacement sensor and a structured light scanner. At the same time, the real-time temperature field data and the real-time surface morphology data of the grinding are preprocessed. Thermodynamic features are extracted from the temperature field data of real-time grinding, and geometric features are extracted from the morphology data of real-time grinding. By integrating thermodynamic and geometric features, a real-time processing state description vector is generated. The real-time machining status description vector is analyzed, and the initial feed compensation parameters are adjusted based on the analysis results.

6. The intelligent compensation control method for precision feed of a linear guide CNC grinding machine according to claim 5, characterized in that, The real-time machining status description vector is analyzed, and the initial feed compensation parameters are adjusted based on the analysis results, including: Anomaly vectors are determined based on a preset threshold detection algorithm to identify real-time processing status description vectors. Based on a pre-defined vector-range database, the controllable range corresponding to the abnormal vector of each real-time processing state description vector is determined. If the abnormal vector of the real-time machining status description vector exceeds the controllable range, the grinding path segment where the abnormal vector is located will be marked as an abnormal path. Based on the anomaly vector, the corresponding preset adjustment strategy is called from the preset adjustment strategy database to adjust the initial feed compensation parameters of the anomaly path.

7. The intelligent compensation control method for precision feed of a linear guide CNC grinding machine according to claim 1, characterized in that, A grinding quality evaluation model is established, and closed-loop feedback control is performed using real-time collected post-grinding surface feature data until the preset surface quality indicators are achieved, including: Collect surface temperature field distribution data and surface geometric feature data after grinding, and simultaneously record process parameter data during the processing; The residual stress distribution data of the surface is determined based on the surface temperature field distribution data after grinding, the surface roughness parameters are extracted based on the surface geometric feature data after grinding, and the quality feature correlation matrix is ​​established by combining the process parameter data. Input the quality feature correlation matrix into the pre-trained grinding quality evaluation model, output the quality score distribution map, and determine the comprehensive quality score based on the quality score distribution map; Determine whether the overall quality score meets the quality requirements based on preset standards; For areas in the quality score distribution map that do not meet the quality requirements, a compensation processing path is generated, the process parameter data is adjusted, and an execution command is issued until the preset surface quality index is achieved.

8. The intelligent compensation control method for precision feed of a linear guide CNC grinding machine according to claim 7, characterized in that, Generate compensatory processing paths for areas in the quality score distribution map that do not meet quality requirements, including: Low-scoring regions were identified using a region growing algorithm based on the quality score distribution map. A spatial location mapping of the defect area is established based on the low-scoring area, and a compensation processing path is determined based on the morphological characteristics of the defect area and the preset feature-path database.

Images