Injection mold core machining precision control method and system

Abstract

The invention relates to the technical field of machining precision control, and discloses an injection mold core machining precision control method and system.The method comprises the steps that a multi-point trigger type sensor is adopted for conducting data collection on the surface of a core of an injection mold, and core surface contour data, geometric dimension data and surface roughness data are obtained; inputting a feature extraction model, extracting fusion features of the surface of the mold core, and performing regional processing on the surface of the mold core to obtain regional polishing parameters; normalized error coefficient calculation is carried out, and a core surface quality distribution diagram is generated; calculating the compensation coefficient of each region to obtain a compensation optimization parameter matrix; and multi-axis linkage path planning is conducted, the surface of the mold core is divided into a plurality of grid units, the curvature value and the polishing track compensation value of each grid unit are calculated, a thermal deformation compensation polishing path is generated, and the automation level and the intelligent degree of mold core machining are improved.

Description

Technical Field

[0001] The invention relates to the technical field of machining accuracy control, and in particular to a method and system for controlling machining accuracy of an injection mold core. Background Art

[0002] The injection mold core is a key component in the injection molding process. Its processing accuracy directly affects the quality of the injection molded product and the service life of the mold. With the growing demand for high-precision injection molded products, traditional core processing methods have been unable to meet the processing requirements of high precision and high surface quality. The current core processing accuracy control method mainly relies on manual experience and single process parameter adjustment, lacking a systematic and intelligent control strategy.

[0003] In the process of precision grinding of cores, it is difficult to ensure the machining accuracy due to the complex surface shape of the cores, various machining parameters, and the influence of uncertain factors such as thermal deformation. Traditional grinding process parameter control methods often use fixed process parameters and cannot adjust the machining parameters in real time according to the changes in the surface characteristics of the cores, resulting in problems such as uneven machining accuracy in local areas and unstable surface quality. Summary of the invention

[0004] The invention provides an injection mold core machining accuracy control method and system, and the invention improves the automation level and intelligent degree of core machining.

[0005] In a first aspect, the present invention provides a method for controlling the machining accuracy of an injection mold core, the method comprising: A multi-point trigger sensor is used to collect data on the core surface of the injection mold to obtain the core surface profile data, geometric dimension data and surface roughness data; Inputting the core surface profile data, the geometric dimension data and the surface roughness data into a feature extraction model, extracting the core surface fusion features, and performing regional processing on the core surface to obtain regional grinding parameters; Calculating a normalized error coefficient based on the regional grinding parameters and generating a core surface quality distribution map; Inputting the core surface quality distribution diagram into a dynamic compensation model including a local compensation amount calculation module and a compensation parameter optimization module to calculate the compensation coefficient of each region, and obtaining a compensation optimization parameter matrix; Based on the compensation optimization parameter matrix, multi-axis linkage path planning is performed, the core surface is divided into a plurality of grid units, and the curvature value and grinding trajectory compensation value of each grid unit are calculated respectively to generate a thermal deformation compensation grinding path.

[0006] In a second aspect, the present invention provides an injection mold core machining accuracy control system, the injection mold core machining accuracy control system comprising: The acquisition module is used to acquire data on the core surface of the injection mold using a multi-point trigger sensor to obtain the core surface profile data, geometric dimension data and surface roughness data; A processing module, used for inputting the core surface profile data, the geometric dimension data and the surface roughness data into a feature extraction model, extracting the core surface fusion feature, and performing regional processing on the core surface to obtain regional grinding parameters; A calculation module, used for calculating a normalized error coefficient based on the regional grinding parameters and generating a core surface quality distribution map; An optimization module, used for inputting the core surface quality distribution map into a dynamic compensation model including a local compensation amount calculation module and a compensation parameter optimization module to calculate the compensation coefficient of each region, and obtain a compensation optimization parameter matrix; A generation module is used to perform multi-axis linkage path planning based on the compensation optimization parameter matrix, divide the core surface into multiple grid units, and calculate the curvature value and grinding trajectory compensation value of each grid unit respectively, and generate a thermal deformation compensation grinding path.

[0007] The third aspect of the present invention provides a computer device, comprising: a memory and at least one processor, wherein the memory stores instructions; the at least one processor calls the instructions in the memory so that the computer device executes the above-mentioned injection mold core processing accuracy control method.

[0008] A fourth aspect of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores instructions, which, when executed on a computer, enable the computer to execute the above-mentioned method for controlling the machining accuracy of an injection mold core.

[0009] In the technical solution provided by the present invention, core surface data is collected by multi-point trigger sensors, and data fusion is performed in combination with a feature extraction model, so as to realize accurate identification of core surface features, and adopt a regional grinding parameter control strategy, based on a normalized error coefficient calculation method, to realize accurate processing control of different areas of the core surface, and improve the uniformity of processing accuracy; a dynamic compensation model and a dual particle swarm algorithm are introduced to realize real-time optimization and compensation of processing parameters, and effectively overcome the influence of uncertainty factors in the processing process; through multi-axis linkage path planning and thermal deformation compensation mechanism, the thermal deformation problem in the grinding process is solved, and the stability of processing accuracy is guaranteed; based on the hierarchical processing method of the evaluation detection matrix, accurate evaluation and visual display of the core surface quality are realized, and the accuracy of processing quality control is improved; through the coordinated optimization of algorithms and the dynamic adjustment of parameters, the automation level and intelligence degree of core processing are significantly improved, manual intervention is reduced, and processing efficiency is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0011] Figure 1 Schematic diagram of the steps of the method for controlling the machining accuracy of the injection mold core in an embodiment of the present invention; Figure 2 It is a structural schematic diagram of an injection mold core machining accuracy control system in an embodiment of the present invention; Figure 3 It is a schematic block diagram of the structure of a computer device in an embodiment of the present invention. DETAILED DESCRIPTION

[0012] Embodiments of the present invention provide a method and system for controlling the machining accuracy of an injection mold core. The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "including" or "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product, or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products, or devices.

[0013] For ease of understanding, the specific process of the embodiment of the present invention is described below. Figure 1 , an embodiment of the method for controlling the machining accuracy of the injection mold core in the embodiment of the present invention includes: Step S1, using a multi-point trigger sensor to collect data on the core surface of the injection mold to obtain core surface profile data, geometric dimension data and surface roughness data; It is understandable that the execution subject of the present invention may be an injection mold core machining accuracy control system, or a terminal or a server, which is not specifically limited here. The embodiment of the present invention is described by taking a server as an execution subject as an example.

[0014] Specifically, a multi-point trigger sensor is used to accurately collect data on the core surface of the injection mold. In order to ensure the comprehensiveness and high accuracy of the data, the core surface of the injection mold is divided into 16 measurement areas. A trigger sensor is set in each measurement area to form a 4×4 matrix sensor layout scheme. This layout scheme evenly covers the core surface, so that the collected measurement data can fully reflect the contour, geometric dimensions and surface roughness of the core surface. The sensors in the 4×4 matrix sensor layout scheme are initialized and calibrated. The sensors are preset through the standard calibration procedure to ensure that they can accurately capture the subtle changes on the core surface in the subsequent data collection process. After the calibration is completed, the sensor collects the original measurement data of the core surface according to the preset sampling frequency, covering all dimensions of the core surface, including contour, geometric dimensions and surface roughness information. High-precision laser measurement is performed based on the original measurement data to obtain higher measurement accuracy and richer surface feature information, and the three-dimensional contour data of the core surface is continuously obtained through laser scanning. The data can accurately reflect the slight changes, unevenness and shape deviations on the core surface. The core surface profile data is converted to a coordinate system. The converted data is compatible with the coordinate system of the processing equipment, which is convenient for subsequent geometric dimension calculation and path planning. The converted data is measured point by point in three coordinates using a preset stepping interval to obtain the geometric dimension data of the core. Optical scanning is performed based on the core geometric dimension data to obtain high-resolution surface morphology data, reflecting the microscopic morphology of the core surface. In order to improve the accuracy of the data and reduce noise interference, the original surface morphology data is filtered to obtain the core surface roughness data.

[0015] Step S2, inputting the core surface profile data, geometric dimension data and surface roughness data into the feature extraction model, extracting the core surface fusion features, and performing regional processing on the core surface to obtain regional grinding parameters; Specifically, the core surface contour data, geometric dimension data and surface roughness data are imported into the feature extraction model. The model includes a feature encoding layer and a feature processing layer. The feature encoding layer adopts the U-Net structure, which can effectively capture the local and global features of the data during the processing process, ensuring that different types of data can be fully encoded and understood. At the same time, the feature processing layer adopts the ResNet structure, which can effectively handle complex feature fusion, help the model learn and retain important information, and ensure that the fusion features of the final output have high accuracy and representativeness. In the feature encoding layer, the first feature encoder mainly processes the contour data of the core surface. By performing five layers of convolution operations on these contour data, the LeakyReLU activation function and batch normalization are used after each layer of convolution operations to avoid overfitting and accelerate the convergence process. This processing method helps to improve the expression ability of the core contour features, so that the contour data can more accurately reflect the changes in the microscopic morphology of the core surface. The second feature encoder processes the geometric dimension data. Similarly, these data are subjected to five layers of convolution operations, and the ReLU activation function and batch normalization are used to ensure that the geometric dimension features can be accurately encoded and improve the effect of feature extraction. For the surface roughness data, the third feature encoder in the feature encoding layer is also processed by five layers of convolution operations, and the ReLU activation function and batch normalization technology are also used after each layer of convolution operation to ensure that the details of the surface roughness data are fully captured. After completing this series of feature encoding processes, the core contour feature encoding, geometric dimension feature encoding and surface roughness feature encoding are input into the feature processing layer. The feature processing layer fuses these three different types of features to generate a unified core surface fusion feature. In order to ensure the excellent fusion effect, the feature processing layer uses nine residual blocks, each of which contains two convolutional layers and a short-circuit connection. The design of the residual block can effectively prevent the gradient disappearance problem and ensure that the information flow can pass through multiple levels of the network, retaining the most important feature information at the output. A sub-regional partitioning model is established based on the core surface fusion features. The sub-regional partitioning model adopts the ResNet structure enhanced by the attention mechanism. The core surface is feature partitioned by three sub-models of different scales. The structure of the sub-model includes four convolutional layers and a fully connected layer. Each sub-model processes data at different scales and can capture the feature differences of the core surface at different levels. This step enables the model to not only effectively distinguish the features of different areas on the core surface, but also make finer divisions between the areas to ensure better control of the machining accuracy of each area. Based on the core surface partition results, the grinding parameters of each area corresponding to the machining stage are calculated. The grinding parameters include grinding pressure, grinding speed, and grinding time, which directly affect the machining accuracy and surface quality of each area.The preliminary grinding parameter calculation is based on the characteristic information of each area, such as contour features, geometric size features, and surface roughness features. The initial grinding parameters of each area are optimized with processing accuracy constraints and surface quality constraints. During the optimization process, the processing difficulty, error tolerance, and surface roughness requirements of each area are considered, and the grinding parameters of each area are adjusted through comprehensive analysis. The regional grinding parameters are obtained.

[0016] Step S3, calculating the normalized error coefficient based on the regional grinding parameters, and generating a core surface quality distribution map; Specifically, the grinding pressure, grinding speed and grinding time in the regional grinding parameters are standardized to eliminate the differences in the dimensions and value ranges of different parameters, so that each parameter can be compared and calculated under a unified standard to obtain a standardized grinding parameter matrix. The normalized error coefficient is calculated based on the standardized grinding parameter matrix to quantify the errors generated in different regions during the processing. By calculating the error coefficient of each region, it is clear which regions have large processing accuracy deviations, so as to make targeted optimization adjustments. The regional error coefficient is calculated by difference with the preset core surface standard parameters to obtain error deviation data, which shows the difference between each region and the standard parameters, thereby helping to evaluate the processing quality of the entire core surface. Based on the error deviation data and the regional error coefficient, an evaluation detection matrix is ​​constructed to evaluate the error of each detection unit. The evaluation detection matrix can provide a detailed framework for subsequent error analysis, and each detection unit represents the processing quality of a specific area on the core surface. In the evaluation detection matrix, the weighted error calculation method is used to weight the error of each detection unit according to the error coefficient of different regions to obtain the detection unit evaluation data. The evaluation data of the inspection units are graded, and the quality of each inspection unit is divided into different grades to obtain the quality grade distribution of the core surface. According to the quality grade distribution, the error grade is visualized to generate surface quality distribution data that is easier to understand. The visualization of the quality grade clearly displays the areas of different quality grades through color identification or heat map. The boundaries of adjacent inspection units in the surface quality distribution data are linearly transitioned. Since there will be a certain transition effect in the error difference between adjacent areas, these differences can be smoothed by linear transition, making the error distribution more continuous. Calculating the error coefficient of the transition area can effectively avoid sudden changes or discontinuities during the processing process, ensuring that the entire surface quality distribution diagram is smooth and in line with the actual processing characteristics. Finally, a continuous core surface quality distribution diagram is obtained.

[0017] Step S4, inputting the core surface quality distribution diagram into a dynamic compensation model including a local compensation amount calculation module and a compensation parameter optimization module to calculate the compensation coefficient of each region, and obtaining a compensation optimization parameter matrix; Specifically, multiple compensation areas are constructed based on the core surface quality distribution map. Each compensation area represents an area on the core surface that needs to be adjusted for precision, and the area is divided based on the quality grade data of the quality distribution map. On the basis of the compensation area, a local compensation amount calculation module is established, and the task of the module is to calculate the pressure compensation coefficient, speed compensation coefficient and time compensation coefficient of each compensation area. These coefficients are used to accurately compensate the processing process of each area to ensure that each area can achieve the predetermined processing accuracy. Through the calculation of the local compensation amount, a regional compensation coefficient matrix is ​​obtained. According to the regional compensation coefficient matrix and the quality grade distribution, the grinding pressure deviation, grinding speed deviation and grinding time deviation of each compensation area are calculated. By calculating these deviation values, a reference value of the compensation parameter is established for each compensation area. The role of the compensation parameter reference value is to provide a basis for the establishment of a dynamic compensation model to ensure that the compensation process can be effectively corrected according to the errors in actual processing. Based on the compensation parameter reference value, a dynamic compensation model is established. The dynamic compensation model calculates a first compensation parameter set according to the reference value, which includes preliminary correction parameters for each compensation area. Through this model, dynamic adjustment of the core surface quality is achieved to adapt to the processing errors in different areas. The first compensation parameter set is subjected to parameter constraint judgment to ensure that each compensation parameter meets the requirements of the processing technology and is within a reasonable error range to obtain the second compensation parameter set, and the second compensation parameter set is subjected to weighted operation with the area weight coefficient of each compensation area. The relative importance of the compensation parameters is adjusted according to the area size of each area to ensure that the correction amount of each area during the compensation process is proportional to its area, thereby improving the overall compensation accuracy. The third compensation parameter set is obtained through weighted operation. The third compensation parameter set is iteratively optimized through the dual particle swarm algorithm in the compensation parameter optimization module. The dual particle swarm algorithm is an optimization algorithm based on swarm intelligence that can find the optimal solution through multiple iterations. In this optimization process, the algorithm continuously adjusts the compensation parameter set to find the target compensation parameter set that can minimize the error. According to the specific conditions of multiple compensation areas, the target compensation parameter set is reconstructed into a compensation optimization parameter matrix to represent the compensation optimization parameters of each area.

[0018] The third compensation parameter set is initialized by the dual particle swarm algorithm in the compensation parameter optimization module. Two particle swarms are constructed, namely the first swarm and the second swarm. Each swarm contains 40 particles, and each particle represents the pressure compensation ΔP, the speed compensation ΔV and the time compensation ΔT in the three-dimensional parameter space, which respectively represent the compensation adjustments required for the core in different processing processes. The particle dimension design enables each particle to contain all the compensations that need to be adjusted. The fitness is calculated for each particle in the first swarm and the second swarm. The fitness function is used to evaluate the quality of each particle. The fitness function is defined as f(x)=w1×|ΔP|+w2×|ΔV|+w3×|ΔT|, where f(x) represents the fitness value, w1, w2 and w3 are the weight coefficients of pressure compensation, speed compensation and time compensation, and the weight values ​​are 0.4, 0.3 and 0.3 respectively. The setting of these weight coefficients reflects the relative importance of each compensation in the final optimization process. The fitness function is calculated by weighted summing the absolute values ​​of ΔP, ΔV and ΔT of each particle to obtain the fitness score of each particle. The first population and the second population obtain their own fitness evaluation results through the fitness function. According to the fitness evaluation results, the two populations are sorted. Based on the fitness evaluation results of the first population, the particles are sorted to obtain the first optimization initial value; similarly, based on the fitness evaluation results of the second population, the second population is sorted to obtain the second optimization initial value. These initial values ​​serve as the starting point for particle swarm optimization. The particles are updated with velocity. Based on the first optimization initial value, the velocity of the particles of the first population is updated to obtain the new velocity of the first particle; similarly, based on the second optimization initial value, the velocity of the second population is updated to obtain the new velocity of the second particle. The velocity of the particle is adjusted so that it can approach the optimal solution faster. The new velocity of the particle will affect its position change. According to the new velocity of the first particle and the new velocity of the second particle, the position update calculation is performed to obtain the updated position of the particle. After the position is updated, the new position of the particle is processed by boundary. Since the calculation range of compensation parameters is limited, the update positions of some particles will exceed these limits. In order to ensure the reasonable position of particles, the parameters that exceed the compensation range are limited to a reasonable range to obtain the first effective compensation parameter and the second effective compensation parameter. Based on the first effective compensation parameter and the second effective compensation parameter, a population crossover operation is performed. The diversity of the particle swarm is improved through information exchange, which accelerates the search process and avoids local optimal solutions. High-quality particles will exchange information through arithmetic crossover to generate a new set of target compensation parameters. The arithmetic crossover method enables the particles after the crossover to contain the characteristics from the parent particles, so that the particles can develop towards a better solution space. The generated target compensation parameter set is reorganized according to the compensation area to construct a compensation optimization parameter matrix.The dimension of the matrix is ​​n×m, where n is the number of compensation areas and m is the dimension of the compensation parameter, which is usually 3, corresponding to the pressure compensation, speed compensation and time compensation respectively.

[0019] Step S5: Perform multi-axis linkage path planning based on the compensation optimization parameter matrix, divide the core surface into multiple grid units, and calculate the curvature value and grinding trajectory compensation value of each grid unit respectively to generate a thermal deformation compensation grinding path.

[0020] Specifically, the core surface is meshed based on the compensation optimization parameter matrix. The core surface is divided into an 8×8 uniform grid unit group, and each grid unit represents a local area on the core surface. Each grid unit contains three compensation parameters: pressure compensation ΔP, speed compensation ΔV, and time compensation ΔT. The local curvature value of each grid unit is calculated. The local curvature value describes the degree of curvature of a certain area on the core surface. According to the calculated local curvature value, the pressure compensation of each grid unit is corrected. The curvature change of the surface will directly affect the distribution of pressure during the grinding process. Areas with larger curvature require higher pressure to achieve the expected processing effect. By correcting the pressure compensation, the grinding process of each grid unit can be controlled more accurately. The speed compensation of grinding is adjusted in linkage. The speed adjustment is closely related to the pressure compensation. The corrected pressure compensation will affect the optimal setting of the grinding speed, thereby ensuring the stability and accuracy of the entire grinding process. The thermal effect analysis of the corrected pressure compensation and the corrected speed compensation is carried out. During the high-precision grinding process, the accumulation of heat will cause thermal deformation on the core surface, which will affect the processing accuracy. By calculating the temperature rise value of each grid unit, the thermal deformation of each area during the grinding process is estimated, and the thermal deformation compensation data is obtained accordingly. The thermal deformation compensation data is obtained based on the results of thermal effect analysis and is used to correct the time compensation in the grinding process. The correction of thermal deformation compensation is to accurately adjust the time under the influence of thermal effect, so as to avoid the processing error caused by temperature change and ensure that the processing accuracy of each grid unit is always kept within the predetermined range. The corrected pressure compensation, corrected speed compensation and corrected time compensation are used for multi-axis linkage path planning to determine the motion trajectory of each grid unit on the entire core surface. Due to the different compensation requirements of different grid units, the path of the grinding head is dynamically adjusted according to the compensation parameters of each grid unit to ensure that each area can be reasonably polished. By calculating the motion trajectory of each grid unit, a series of grinding path points are obtained. In order to ensure smoothness and accuracy during the grinding process, the grinding path points are smoothed to reduce the uneven processing caused by path mutations, thereby improving the stability and accuracy of the processing process. By smoothing the trajectory of the grinding path point sequence, a thermal deformation compensation grinding path is obtained.

[0021] In the embodiment of the present invention, core surface data is collected by multi-point trigger sensors, and data fusion is performed in combination with a feature extraction model to achieve accurate identification of core surface features. A regional grinding parameter control strategy is adopted, and based on the normalized error coefficient calculation method, accurate processing control of different areas of the core surface is achieved, and the uniformity of processing accuracy is improved; a dynamic compensation model and a dual particle swarm algorithm are introduced to achieve real-time optimization and compensation of processing parameters, and effectively overcome the influence of uncertainty factors in the processing process; through multi-axis linkage path planning and thermal deformation compensation mechanism, the thermal deformation problem in the grinding process is solved, and the stability of processing accuracy is guaranteed; based on the hierarchical processing method of the evaluation detection matrix, accurate evaluation and visual display of the core surface quality are achieved, and the accuracy of processing quality control is improved; through the collaborative optimization of algorithms and the dynamic adjustment of parameters, the automation level and intelligence degree of core processing are significantly improved, manual intervention is reduced, and processing efficiency is improved.

[0022] In a specific embodiment, the process of executing step S1 may specifically include the following steps: The core surface of the injection mold is divided into 16 measurement areas by using multi-point trigger sensors. A trigger sensor is set for each measurement area to obtain a 4×4 matrix sensor layout scheme. Initialize and calibrate the sensors in the 4×4 matrix sensor arrangement scheme, collect data according to the preset sampling frequency, and obtain the original measurement data of the core surface; Based on the original measurement data, high-precision laser measurement is performed to obtain the three-dimensional contour features of the core surface through continuous scanning to obtain the core surface contour data; The core surface profile data is converted into a coordinate system to obtain converted data, and the converted data is measured point by point in three coordinates using a preset stepping interval to obtain core geometric dimension data; Based on the core geometric size data, the core surface is optically scanned to obtain original surface morphology data, and the original surface morphology data is filtered to obtain the core surface roughness data.

[0023] Specifically, the core surface is divided into multiple measurement areas through sensor arrangement. The core surface is divided into 16 measurement areas, so that the accuracy of each area is effectively controlled. By setting a trigger sensor in each area, a 4×4 matrix sensor arrangement scheme is constructed, and each sensor is responsible for the measurement task in the corresponding area. The sensor is initialized and calibrated. The reliability of the measurement data is ensured by adjusting the measurement position and acquisition parameters of the sensor. During the calibration process, a standard workpiece or a reference object of known size is used for verification, and the sensor deviation is corrected by the calibration model. After the calibration is completed, the sensor collects data on the core surface according to the preset sampling frequency. The selection of the sampling frequency needs to be determined according to the complexity and accuracy requirements of the core surface to ensure that enough measurement points are collected to provide high-precision data support. Through this step, the original measurement data of the core surface is obtained. High-precision laser measurement is performed based on the original measurement data to improve the measurement accuracy. Laser measurement is a high-precision surface detection method that obtains its three-dimensional contour features by continuously scanning the core surface. The advantage of laser scanning is that it can obtain surface height change information and is not affected by the surface material or color. Through this step, the detailed three-dimensional contour data of the core surface is obtained, including the spatial coordinate information of each measurement point. Assume that the coordinates of each measurement point in the original measurement data are ,in represents the index of the measurement point, , , are the positions of the point on the three coordinate axes respectively. These coordinate values ​​are obtained by laser scanning the core surface, and a complete three-dimensional model of the core surface is constructed. After obtaining the three-dimensional contour data of the core surface, the coordinate system is transformed. Since the coordinate system used in the measurement process is different from the actual processing environment, the coordinate system of the measurement data is transformed to adjust it to the standard processing coordinate system. Assume that the coordinates of the original measurement data in the sensor coordinate system are , after the coordinate system transformation, the transformed data is ,in, is the converted data. Common methods of coordinate system transformation include translation transformation, rotation transformation and scaling transformation. The specific transformation matrix is ​​determined according to the layout of the sensor and the actual posture of the workpiece. The converted data is measured point by point in three coordinates to obtain the geometric dimension data of the core. Point-by-point measurement is performed using a preset stepping spacing, that is, the coordinates of each measuring point are collected one by one according to a certain distance step length. In this measurement method, the interval between each measuring point remains consistent and is set by the stepping spacing. For example, if the stepping spacing is , then in the transformed data, the interval of the measurement points is Millimeters. The geometric dimension data of each measuring point is obtained by calculating its spatial position. Assuming that the position of each measuring point is , the geometric dimension data of the core is obtained by calculating the three-dimensional coordinate value of each point, thereby obtaining the complete core geometric information. Based on the core geometric dimension data, optical scanning is performed to obtain the surface morphology data of the core surface. Optical scanning obtains the tiny morphological features of the core surface through reflected light or refracted light technology, which is suitable for capturing surface texture and details. The original surface morphology data obtained by optical scanning contains the surface height information of each measurement point. If it is assumed that the core surface morphology data is (i.e. The surface height of each measurement point) is then combined with the geometric dimension data to form complete surface topography information. The original surface topography data is filtered to improve the accuracy and reliability of the data. Different types of filtering algorithms are used in the filtering process, such as mean filtering, median filtering or Gaussian filtering. For example, for each surface height data , apply a window function to smooth it, the formula is as follows: ; in, is the filtered data, For the The surface height of the original data points, is the size of the filter window. Through filtering, noise is effectively removed and surface topography data is smoothed to obtain more accurate core surface roughness data.

[0024] In a specific embodiment, the process of executing step S2 may specifically include the following steps: The core surface profile data, geometric dimension data and surface roughness data are imported into the feature extraction model. The feature extraction model includes a feature encoding layer and a feature processing layer. The feature encoding layer adopts a U-Net structure, and the feature processing layer adopts a ResNet structure. The first feature encoder in the feature encoding layer performs 5 layers of convolution operations on the core surface contour data. After each layer of convolution operation, LeakyReLU activation function and batch normalization are used to obtain the core contour feature encoding; The second feature encoder in the feature encoding layer performs 5 layers of convolution operations on the geometric size data, and uses the ReLU activation function and batch normalization after each layer of convolution operation to obtain the geometric size feature encoding; For the third feature encoder in the feature encoding layer, five layers of convolution operations are performed on the surface roughness data. After each layer of convolution operation, the ReLU activation function and batch normalization are used to obtain the surface roughness feature encoding; The core contour feature coding, geometric dimension feature coding and surface roughness feature coding are input into the feature processing layer for feature fusion to obtain the core surface fusion feature. The feature processing layer contains 9 residual blocks, each of which contains two convolutional layers and a short-circuit connection. A regional partitioning model is established based on the fusion features of the core surface. The regional partitioning model adopts the ResNet structure enhanced by the attention mechanism. The core surface is partitioned by three sub-models of different scales. Each sub-model contains four convolutional layers and one fully connected layer to obtain the core surface partitioning result. Based on the core surface zoning results, the grinding pressure, grinding speed and grinding time of the corresponding processing stage are calculated respectively to obtain the initial grinding parameters of each area. The initial grinding parameters of each area are optimized with processing accuracy constraints and surface quality constraints to obtain the regional grinding parameters.

[0025] Specifically, the core surface contour data, geometric dimension data and surface roughness data are feature extracted and analyzed to extract effective feature information from the original data. The feature extraction model consists of two main parts: a feature encoding layer and a feature processing layer. In the design of the feature encoding layer, the U-Net structure is used to effectively encode the core surface contour data, geometric dimension data and surface roughness data. The U-Net structure is a neural network structure used for image processing, which can effectively extract the spatial features of the input data through convolution operations. For the core surface contour data, the first feature encoder is used for processing. The core surface contour data is feature extracted through 5 layers of convolution operations. After each layer of convolution, the LeakyReLU activation function is used for nonlinear transformation. The form of the LeakyReLU function is: ; in, is a small constant, set to 0.01. Batch Normalization is used to normalize the output to reduce the variation of each layer input, thereby accelerating training and improving the stability of the model. The feature map after each layer of convolution is LeakyReLU activated and batch normalized to obtain the core surface contour feature encoding. Similarly, the second feature encoder processes the geometric dimension data. The feature extraction of geometric dimension data is also performed through 5 layers of convolution operations, and each layer of convolution is processed using the ReLU activation function and batch normalization. The ReLU activation function is defined as: ; The ReLU function has the characteristics of simple calculation, can quickly perform activation operations and effectively avoid the problem of gradient disappearance. The third feature encoder performs similar convolution processing on the surface roughness data. The roughness data is also processed by 5 layers of convolution and ReLU activation function to obtain the surface roughness feature encoding. These feature codes correspond to different aspects of the core: contour, geometric size and surface roughness. The core surface contour feature encoding, geometric size feature encoding and surface roughness feature encoding are input into the feature processing layer for feature fusion. In the feature processing layer, the ResNet structure is used for feature fusion. ResNet (residual network) is a deep neural network structure that solves the gradient disappearance problem in deep networks by introducing residual connections. The feature processing layer consists of multiple residual blocks, each of which contains two convolutional layers and a short-circuit connection. Based on the residual block, the network can automatically learn the complex relationship between features and effectively fuse the contour features, geometric size features and surface roughness features into a unified core surface fusion feature. Assume that the input feature of the residual block is , after convolution and activation function processing, the output is , then the output of the residual block is: ; The residual learning method enables the network to learn deeper features while avoiding the problem of gradient disappearance. The output of the feature processing layer is the core surface fusion feature, which integrates all the important information of the core surface. The core surface is partitioned using a feature-based region partitioning model. The model adopts a ResNet structure enhanced by an attention mechanism, which can automatically learn the most appropriate partitioning method by focusing on the feature differences of different regions on the core surface. In order to adapt to the characteristics of different regions, the model uses three sub-models of different scales, each of which contains 4 convolutional layers and a fully connected layer. By analyzing the surface features of the core, the sub-model can carefully divide the core surface into several regions. These regions will correspond to different processing requirements in the subsequent processing process, such as different regions require different grinding pressures, speeds and times. Based on the output of the sub-model, the partitioning results of the core surface are obtained. Based on the regional information after partitioning, the grinding pressure, grinding speed and grinding time of the corresponding processing stage are calculated according to the characteristics of each region. These initial grinding parameters can reflect the different requirements of different regions for accuracy and surface quality. The initial grinding parameters of each region are optimized for processing accuracy constraints and surface quality constraints. During the optimization process, certain constraints are used, such as the maximum and minimum limits of grinding pressure, speed and time, to ensure that the parameters do not exceed the acceptable range during the actual processing. Through the optimization operation, the final regional grinding parameters can effectively ensure the processing accuracy and surface quality of each area.

[0026] In a specific embodiment, the process of executing step S3 may specifically include the following steps: The grinding pressure, grinding speed and grinding time in the regional grinding parameters are standardized respectively to construct a standardized grinding parameter matrix; Calculate the normalized error coefficient based on the standardized grinding parameter matrix to obtain the regional error coefficient, and perform difference calculation between the regional error coefficient and the preset core surface standard parameter to obtain the error deviation data; Based on the error deviation data and the regional error coefficient, an evaluation detection matrix is ​​constructed, and a weighted error calculation is performed on each detection unit to obtain the detection unit evaluation data; The evaluation data of the detection unit is graded to obtain the quality grade distribution, and the error grade is visualized according to the quality grade distribution to obtain the surface quality distribution data; Linear transition is performed on the boundaries of adjacent detection units in the surface quality distribution data, the error coefficient of the transition area is calculated, the continuous distribution data is obtained, and the core surface quality distribution map is generated based on the continuous distribution data.

[0027] Specifically, the grinding pressure, grinding speed and grinding time in the grinding parameters are standardized to eliminate the dimensional differences between different parameters so that they can be compared and optimized under the same standard. The Z-score standardization method is used to transform each parameter into a standard normal distribution. The specific formula is: ; in, is the original grinding parameter, and are the mean and standard deviation of the parameter, respectively. is the standardized parameter. The grinding pressure, grinding speed and grinding time are converted into standardized data with zero mean and unit variance, and a standardized grinding parameter matrix is ​​constructed. Based on the standardized grinding parameter matrix, the normalized error coefficient is calculated, and the processing accuracy of each area is evaluated by the coefficient. Normalized error coefficient Reflects the deviation between the actual grinding parameters and the preset standard parameters. Assume that the preset core surface standard parameters are ,in The standard grinding pressure, speed and time are respectively. For each area , the normalized error coefficient is calculated by the following formula: ; in, For Region The actual grinding parameters, are the preset standard parameters. It represents the absolute deviation between the actual parameter and the standard parameter, and is divided by the standard parameter The error coefficient has the same dimension in different units. The normalized error coefficient is calculated The degree of deviation in each area can be measured. The difference calculation is performed using the regional error coefficient to obtain the error deviation data. Assuming the regional error coefficient is a 1 vector, indicating the degree of error in each area, and the preset standard parameters is a constant matrix, error deviation data Calculated by: ; This formula represents the error deviation data obtained by difference calculation , which reflects the gap between the actual processing and the standard processing in each area. Based on the error deviation data and the regional error coefficient, an evaluation detection matrix is ​​constructed, and a weighted error calculation is performed on each detection unit. Assuming the error deviation data and the regional error coefficient The evaluation detection matrix is ​​composed of , then the weighted error of each detection unit is calculated by the following formula: ; in, and is a weighted coefficient, which indicates the relative importance of the error deviation and regional error coefficient in the calculation, and is adjusted according to specific processing requirements. It can reflect the error degree of each detection unit. The evaluation data of the detection unit is graded to obtain the quality grade of each area. It will be divided into multiple levels, for example, by setting thresholds , divide the evaluation data into several levels, assuming that the score of each level is , then the quality grade distribution is expressed as follows: ; Grading helps to classify the various areas of the core surface according to the quality level, making it easier to understand and optimize the machining accuracy of each area. According to the quality level distribution, the error level is visualized to generate surface quality distribution data. The quality levels of different areas are displayed using heat maps and other methods. Considering the complexity in actual processing, the boundaries of adjacent inspection units are linearly transitioned, and the error coefficient of the transition area is calculated to obtain a smoother and continuous surface quality distribution. Assuming that there is a boundary transition zone between adjacent areas, the error coefficient of the transition zone Calculated by linear interpolation: ; in, and are the error coefficients of adjacent regions, and are the starting and ending positions of the transition zone, is the position of any point in the transition zone. Through linear transition, a smoother error distribution is obtained, making the mass distribution of the entire core surface more continuous.

[0028] In a specific embodiment, the process of executing step S4 may specifically include the following steps: Based on the core surface quality distribution map, multiple compensation areas are constructed, and a local compensation amount calculation module is established based on the multiple compensation areas to calculate the pressure compensation coefficient, speed compensation coefficient, and time compensation coefficient of each compensation area to obtain a regional compensation coefficient matrix; According to the regional compensation coefficient matrix and the quality grade distribution, the grinding pressure deviation, grinding speed deviation and grinding time deviation of each compensation area are calculated to obtain the compensation parameter reference value; Establishing a dynamic compensation model based on the compensation parameter reference value, and calculating a first compensation parameter set through the dynamic compensation model; Performing parameter constraint judgment on the first compensation parameter set to obtain a second compensation parameter set, and performing weighted operation on the second compensation parameter set and the area weight coefficient of each compensation area to obtain a third compensation parameter set; The third compensation parameter set is iteratively optimized by using a dual particle swarm algorithm in the compensation parameter optimization module to obtain a target compensation parameter set, and the target compensation parameter set is reconstructed into a compensation optimization parameter matrix according to multiple compensation regions.

[0029] Specifically, the core surface is divided by the mass distribution diagram to form multiple compensation areas. Assume that the mass distribution diagram of the core surface is ,in Indicates The quality evaluation results of each area. On the quality distribution map, the core surface is divided into multiple compensation areas for different quality grade areas, and each area needs to be compensated separately. The quality distribution map is segmented by a certain algorithm. Assume that the multiple compensation areas generated are ,in Indicates the number of compensation areas. Based on these compensation areas, a local compensation calculation module is constructed to calculate the grinding compensation coefficient of each area, including the pressure compensation coefficient. , speed compensation coefficient and time compensation factor To ensure that the grinding effect of each area meets the predetermined accuracy requirements. Assume that the pressure, speed and time compensation of each area are calculated to obtain the compensation coefficient matrix. The calculation of the compensation coefficient is based on the required processing accuracy, quality distribution and processing difficulty of different areas. Based on the regional compensation coefficient matrix and quality grade distribution , calculate the grinding pressure deviation, grinding speed deviation and grinding time deviation of each compensation area, and obtain the reference value of the compensation parameter. The quality level is , and its standard grinding parameters are , ,and , then the grinding pressure deviation, speed deviation and time deviation of each area are calculated by the following formula: ; ; ; in, , ,and Respectively represent the area Actual measured grinding pressure, speed and time , ,and is the preset standard parameter. Through the above calculation, the grinding deviation data of each area is obtained. A dynamic compensation model is established based on the compensation parameter benchmark value. This model eliminates the deviation by adjusting the parameters in real time to ensure that the accuracy of each area during the processing meets the requirements. Assume that the dynamic compensation model is , the first set of compensation parameters is obtained through calculation of the model Corresponding to the first area, other areas obtain the corresponding first compensation parameter set. The first compensation parameter set is calculated based on the dynamic compensation model and the regional compensation reference value, and the preliminary compensation result of each area is obtained through the model. The first compensation parameter set is subjected to parameter constraint judgment to ensure that the compensation parameters do not exceed the predetermined range. Assume that the compensation parameters have a preset constraint range ,and , then through constraint judgment, the second compensation parameter set is obtained , these parameters satisfy the following conditions: ; ; ; The second compensation parameter set is weighted by the area weight coefficient of each compensation area to obtain a third compensation parameter set. Assume that each compensation area The area weight coefficient is , the third compensation parameter set is calculated by the following formula: ; in, For Region The area weight coefficient indicates the importance of different areas in the overall compensation. The weighted third compensation parameter set The final compensation optimization parameters are used. The dual particle swarm algorithm in the compensation parameter optimization module is used to iteratively optimize the compensation parameters to improve the compensation accuracy. The dual particle swarm algorithm optimizes the compensation parameter set by simulating the search process of multiple particles in the solution space. Assume that the particle speed is , the location is , the fitness is , in each iteration, the velocity and position of the particle are updated according to the following formula: ; ; in, is the inertia weight, and is the acceleration constant, and is a random number, is the particle's best historical position, is the global optimal position. Through multiple iterations of optimization, the particle swarm algorithm obtains the optimal compensation parameter set , and finally reconstructed into the compensation optimization parameter matrix The final compensation optimization parameter matrix The most optimized grinding parameters will be provided for each compensation area to ensure that the machining accuracy and quality of the entire core surface meet the expected goals.

[0030] In a specific embodiment, the execution step performs compensation parameter iterative optimization on the third compensation parameter set by using a dual particle swarm algorithm in the compensation parameter optimization module to obtain a target compensation parameter set, and reconstructs the target compensation parameter set into a compensation optimization parameter matrix according to multiple compensation regions. The process may specifically include the following steps: The double particle swarm algorithm in the compensation parameter optimization module is used to perform double particle swarm initialization processing on the third compensation parameter set to construct a first population and a second population, each population contains 40 particles, and each particle contains parameters of three dimensions: pressure compensation ΔP, speed compensation ΔV, and time compensation ΔT; The fitness of the first population and the second population is calculated respectively, and the fitness function is f(x)=w1×|ΔP|+w2×|ΔV|+w3×|ΔT|, where f(x) is the fitness function, w1, w2, w3 are the pressure compensation weight coefficient, speed compensation weight coefficient and time compensation weight coefficient, and their values ​​are 0.4, 0.3, 0.3 respectively, ΔP, ΔV, ΔT are the pressure compensation amount, speed compensation amount and time compensation amount respectively, and the first fitness evaluation result of the first population and the second fitness evaluation result of the second population are obtained; The first population is sorted based on the first fitness evaluation result to obtain a first optimized initial value, and the second population is sorted based on the second fitness evaluation result to obtain a second optimized initial value; Performing a velocity update calculation based on the first optimized initial value to obtain a new velocity of the first particle, and performing a velocity update calculation based on the second optimized initial value to obtain a new velocity of the second particle; Perform position update calculation according to the new velocity of the first particle to obtain the new position of the first particle, and perform position update calculation according to the new velocity of the second particle to obtain the new position of the second particle; Perform boundary processing on the first particle new position and the second particle new position respectively, limit the parameters beyond the compensation range to a reasonable range, and obtain the first effective compensation parameter and the second effective compensation parameter; Based on the first effective compensation parameter and the second effective compensation parameter, a population crossover operation is performed, information is exchanged between high-quality particles in the first population and the second population, and a target compensation parameter set is generated by using an arithmetic crossover method; The target compensation parameter set is reorganized according to the compensation area to construct an n×m dimensional compensation parameter matrix, where n is the number of compensation areas and is equal to the number of multiple compensation areas, and m is the compensation parameter dimension and is equal to 3, corresponding to the pressure compensation amount, speed compensation amount, and time compensation amount, respectively, to obtain the compensation optimization parameter matrix.

[0031] Specifically, the dual particle swarm algorithm in the compensation parameter optimization module is used to initialize the third compensation parameter set. The compensation parameter set is divided into two parts to construct two populations. Each population contains 40 particles, each particle represents a solution for a compensation parameter, and is composed of parameters in three dimensions: pressure compensation amount , speed compensation , and time compensation The initialization of these particles is usually randomly selected to ensure the diversity of the search space. Each initialized particle is represented as a three-dimensional vector: ; in It is The compensation amount of each particle in the pressure direction, is the compensation amount in the speed direction, is the compensation amount in the time direction. According to the above initialization rules, create the first population and the second population , and assign a fitness value to each particle. The fitness value is the key to evaluating the quality of particles and is usually calculated through a fitness function. The goal of the fitness function is to comprehensively consider the three compensation amounts of each particle, and to minimize the absolute value of the compensation amount to ensure the accuracy of the processing process. The fitness function is as follows: ; in represents the fitness of the particle, ,and are the weight coefficients of pressure compensation, speed compensation and time compensation respectively. According to the given task requirements, the weight coefficients are set as , , . The fitness of each particle is calculated through this function, and then the population is sorted and the most fit particle is found. The two populations are sorted based on the fitness results. For the first population, its particles are sorted based on the first fitness evaluation result to obtain the first optimized initial value set Similarly, for the second population, we get the second optimized initial value set The particle speed is updated based on the optimized initial value. The particle speed update formula is: ; in Represents particles At the next generation speed, is the current speed, It is a particle The best position in its history, is the global optimal position, and is the acceleration constant, and is a random number. For example, suppose , 1.5, , through these parameters, the particle speed will be updated so that they can move towards the optimal solution. After updating the speed, the particle position is updated. The position update formula is as follows: ; The updated particle position is the new compensation value These new positions are bounded to ensure that the compensation does not exceed a reasonable range. For example, suppose the reasonable range of pressure compensation is Similarly, the speed and time compensation also have corresponding minimum and maximum values. If the updated particle position exceeds these ranges, it needs to be corrected to limit it to a reasonable range. Assume that the corrected compensation is , these values ​​will be used as effective compensation parameters. A population crossover operation is performed based on the first effective compensation parameter and the second effective compensation parameter to enhance the algorithm's exploration ability. The crossover operation is performed in an arithmetic crossover manner, thereby exchanging the solutions of high-quality particles between the two populations. Assume that high-quality particles are selected from the first population. and the high-quality particles in the second population , the target compensation parameter set is obtained by arithmetic crossover , the crossover formula is as follows: ; in, is the crossover weight coefficient, which is usually set to 0.5, indicating that the contributions of two particles in generating new particles are equal. Through the crossover operation, particles with better characteristics are generated, which helps the algorithm converge to the optimal solution faster. The target compensation parameter set obtained through the crossover operation Reorganize according to multiple compensation areas, build dimensional compensation parameter matrix, where is the number of compensation regions, and Corresponding to three compensation parameters: pressure compensation amount , speed compensation , and time compensation .

[0032] In a specific embodiment, the process of executing step S5 may specifically include the following steps: The core surface is meshed based on the compensation optimization parameter matrix to construct an 8×8 uniform grid unit group. The compensation parameters of each grid unit include pressure compensation ΔP, velocity compensation ΔV and time compensation ΔT, and gridded compensation data is obtained. Calculating a local curvature value for each grid unit in the gridded compensation data, correcting a grinding pressure compensation amount according to the local curvature value to obtain a corrected pressure compensation amount, and adjusting a grinding speed compensation amount in linkage based on the corrected pressure compensation amount to obtain a corrected speed compensation amount; Perform thermal effect analysis on the corrected pressure compensation amount and the corrected speed compensation amount, calculate the temperature rise value of each grid unit, obtain thermal deformation compensation data, and correct the grinding time compensation amount based on the thermal deformation compensation data to obtain the corrected time compensation amount; The corrected pressure compensation, speed compensation and time compensation are used for multi-axis linkage path planning, and the motion trajectory of the grinding head in each grid unit is calculated to obtain the grinding path point sequence. The grinding path point sequence is then smoothed to obtain the thermal deformation compensation grinding path.

[0033] Specifically, the core surface is meshed based on the compensation optimization parameter matrix. The matrix contains the parameters of different compensation areas. Based on these parameters, the core surface is divided into multiple grid cells, which are divided into a uniform 8×8 grid cell group. Each grid cell contains three compensation parameters related to the area: pressure compensation amount , speed compensation , and time compensation These parameters are directly extracted from the compensation optimization parameter matrix and mapped to each grid cell on the surface by a simple gridding method to obtain gridded compensation data. Based on each grid cell in the gridded compensation data, the local curvature value of the area is calculated. The local curvature value is estimated by the geometric features of the core surface at this grid cell. Assuming that the precise surface contour of the core is obtained by laser scanning or 3D reconstruction technology, the local curvature In the grid cell The calculation formula at is: ; in is the height of the core surface, and are the horizontal and vertical coordinates respectively. By calculating the curvature of each grid cell, the curvature of the surface in different areas is revealed. The pressure compensation amount during the grinding process is corrected according to the local curvature value. Assuming that the original pressure compensation amount is , then the corrected pressure compensation amount Calculated by the following relationship: ; in, It is a correction coefficient, which is adjusted according to the actual process. By correcting the pressure compensation, it can better adapt to the local geometric changes of the surface, thereby improving the processing accuracy. Based on the corrected pressure compensation, the grinding speed compensation is adjusted in conjunction . Set a linkage relationship, assuming that the correction formula of the speed compensation is: ; in is the linkage correction coefficient, which reflects the mutual influence between pressure and speed. Through this formula, the corrected speed compensation can be coordinated with the pressure compensation to avoid improper speed adjustment due to excessive or insufficient pressure, ensuring efficiency and stability during the grinding process. The thermal effect analysis is performed on the corrected pressure compensation and the corrected speed compensation. The thermal effect is an important factor in injection mold processing, because friction and pressure during the grinding process will cause the core surface temperature to rise, thereby causing thermal deformation. In order to quantify this effect, the temperature rise of each grid cell is calculated using the following thermal effect model : ; in, The heat generated by friction during the grinding process. is the material mass per grid cell, is the specific heat capacity of the material. According to the temperature rise of each grid unit, the thermal deformation compensation data is obtained. Thermal deformation compensation amount It will affect the time compensation of grinding. Through the thermal effect correction formula, the corrected time compensation It is expressed as: ; in is the thermal effect correction coefficient. Through correction, the processing error caused by temperature change is effectively reduced and the stability of the core surface quality is improved. The corrected pressure compensation, speed compensation and time compensation are used for multi-axis linkage path planning. The correction parameters of each grid unit are used to calculate the motion trajectory of the grinding head in this area. Assume that the grinding head is The motion trajectory within a grid cell Described by the following path planning model: ; The specific path planning method is implemented through a multi-axis linkage algorithm, such as differential evolution, particle swarm optimization or other optimization methods. The goal of path planning is to generate a smooth trajectory to ensure that the movement of the grinding head on the core surface is both accurate and efficient. After trajectory smoothing, unnecessary jitter and vibration during the grinding process are eliminated, making the movement more stable and obtaining the final thermal deformation compensation grinding path.

[0034] The above describes the method for controlling the machining accuracy of the injection mold core in the embodiment of the present invention. The following describes the control system for machining accuracy of the injection mold core in the embodiment of the present invention. Figure 2 In one embodiment of the present invention, an injection mold core machining accuracy control system includes: The acquisition module is used to acquire data on the core surface of the injection mold using a multi-point trigger sensor to obtain the core surface profile data, geometric dimension data and surface roughness data; A processing module is used to input the core surface profile data, geometric dimension data and surface roughness data into the feature extraction model, extract the core surface fusion features, and perform regional processing on the core surface to obtain regional grinding parameters; A calculation module is used to calculate the normalized error coefficient based on the regional grinding parameters and generate a core surface quality distribution map; An optimization module is used to input the core surface quality distribution map into a dynamic compensation model including a local compensation amount calculation module and a compensation parameter optimization module to calculate the compensation coefficient of each area and obtain a compensation optimization parameter matrix; The generation module is used to perform multi-axis linkage path planning based on the compensation optimization parameter matrix, divide the core surface into multiple grid units, and calculate the curvature value and grinding trajectory compensation value of each grid unit respectively, and generate a thermal deformation compensation grinding path.

[0035] Through the coordinated cooperation of the above-mentioned components, the core surface data is collected through multi-point trigger sensors, and data fusion is performed in combination with the feature extraction model to achieve accurate identification of the core surface features. The regional grinding parameter control strategy is adopted, and based on the normalized error coefficient calculation method, accurate processing control of different areas of the core surface is achieved, and the uniformity of processing accuracy is improved; the dynamic compensation model and dual particle swarm algorithm are introduced to achieve real-time optimization and compensation of processing parameters, effectively overcoming the influence of uncertain factors in the processing process; through multi-axis linkage path planning and thermal deformation compensation mechanism, the thermal deformation problem in the grinding process is solved, and the stability of processing accuracy is guaranteed; based on the hierarchical processing method of the evaluation detection matrix, accurate evaluation and visual display of the core surface quality are achieved, and the accuracy of processing quality control is improved; through the collaborative optimization of algorithms and dynamic adjustment of parameters, the automation level and intelligence of core processing are significantly improved, human intervention is reduced, and processing efficiency is improved.

[0036] Reference Figure 3 In an embodiment of the present invention, a computer device is also provided. The computer device may be a server, and its internal structure may be as follows: Figure 3As shown. The computer device includes a processor, a memory, a display screen, an input device, a network interface and a database connected through a system bus. Among them, the processor designed by the computer is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store the corresponding data in this embodiment. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, the above method is implemented.

[0037] Those skilled in the art will understand that Figure 3 The structure shown in the figure is merely a block diagram of a portion of the structure related to the solution of the present invention, and does not constitute a limitation on the computer device to which the solution of the present invention is applied.

[0038] An embodiment of the present invention further provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, the above method is implemented. It can be understood that the computer-readable storage medium in this embodiment can be a volatile readable storage medium or a non-volatile readable storage medium.

[0039] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media provided by the present invention and used in the embodiments may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double-speed data rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM.

[0040] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, systems and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0041] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art or the whole or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk and other media that can store program code.

[0042] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features thereof may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for controlling the machining accuracy of an injection mold core, characterized in that: The method comprises: A multi-point trigger sensor is used to collect data on the core surface of the injection mold to obtain the core surface profile data, geometric dimension data and surface roughness data; Inputting the core surface profile data, the geometric dimension data and the surface roughness data into a feature extraction model, extracting the core surface fusion features, and performing regional processing on the core surface to obtain regional grinding parameters; Calculating a normalized error coefficient based on the regional grinding parameters and generating a core surface quality distribution map; Inputting the core surface quality distribution map into a dynamic compensation model including a local compensation amount calculation module and a compensation parameter optimization module to calculate the compensation coefficient of each region, and obtaining a compensation optimization parameter matrix; Based on the compensation optimization parameter matrix, multi-axis linkage path planning is performed, the core surface is divided into a plurality of grid units, and the curvature value and grinding trajectory compensation value of each grid unit are calculated respectively to generate a thermal deformation compensation grinding path.

2. The method for controlling the processing accuracy of the injection mold core according to claim 1, characterized in that: The multi-point trigger sensor is used to collect data on the core surface of the injection mold to obtain the core surface profile data, geometric dimension data and surface roughness data, including: The core surface of the injection mold is divided into 16 measurement areas by using multi-point trigger sensors. A trigger sensor is set for each measurement area to obtain a 4×4 matrix sensor layout scheme. Initializing and calibrating the sensors in the 4×4 matrix sensor arrangement scheme, collecting data at a preset sampling frequency, and obtaining original measurement data of the core surface; Based on the original measurement data, high-precision laser measurement is performed to obtain the three-dimensional contour features of the core surface by continuous scanning to obtain the core surface contour data; Performing coordinate system conversion on the core surface profile data to obtain converted data, and performing point-by-point measurement of three coordinates on the converted data using a preset stepping interval to obtain core geometric dimension data; Based on the core geometric size data, the core surface is optically scanned to obtain original surface morphology data, and the original surface morphology data is filtered to obtain core surface roughness data.

3. The method for controlling the processing accuracy of the injection mold core according to claim 2, characterized in that: The core surface profile data, the geometric dimension data and the surface roughness data are input into a feature extraction model, the core surface fusion feature is extracted, and the core surface is processed by region to obtain the region grinding parameters, including: Importing the core surface contour data, the geometric dimension data and the surface roughness data into a feature extraction model, wherein the feature extraction model comprises a feature encoding layer and a feature processing layer, wherein the feature encoding layer adopts a U-Net structure and the feature processing layer adopts a ResNet structure; The first feature encoder in the feature encoding layer performs 5 layers of convolution operations on the core surface profile data, and uses a LeakyReLU activation function and batch normalization processing after each layer of convolution operations to obtain a core profile feature encoding; The second feature encoder in the feature encoding layer performs 5 layers of convolution operations on the geometric size data, and uses a ReLU activation function and batch normalization after each layer of convolution operations to obtain a geometric size feature encoding; For the third feature encoder in the feature encoding layer, perform 5 layers of convolution operations on the surface roughness data, use a ReLU activation function and batch normalization after each layer of convolution operations to obtain surface roughness feature encoding; Inputting the core contour feature code, the geometric dimension feature code and the surface roughness feature code into the feature processing layer for feature fusion to obtain the core surface fusion feature, wherein the feature processing layer comprises 9 residual blocks, each residual block comprises two convolutional layers and a short circuit connection; A region partitioning model is established according to the core surface fusion features. The region partitioning model adopts a ResNet structure enhanced by an attention mechanism. The core surface is feature partitioned through three sub-models of different scales. Each sub-model includes four convolutional layers and one fully connected layer to obtain a core surface partitioning result. Based on the core surface zoning results, the grinding pressure, grinding speed and grinding time of the corresponding processing stage are calculated respectively to obtain the initial grinding parameters of each area, and the parameters of the initial grinding parameters of each area are optimized under the constraints of processing accuracy and surface quality to obtain the regional grinding parameters.

4. The method for controlling the processing accuracy of the injection mold core according to claim 3, characterized in that: The normalized error coefficient is calculated based on the regional grinding parameters, and a core surface quality distribution diagram is generated, including: The grinding pressure, grinding speed and grinding time in the regional grinding parameters are respectively standardized to construct a standardized grinding parameter matrix; Calculating a normalized error coefficient based on the standardized polishing parameter matrix to obtain a regional error coefficient, and performing difference calculation between the regional error coefficient and a preset core surface standard parameter to obtain error deviation data; Based on the error deviation data and the regional error coefficient, an evaluation detection matrix is ​​constructed, and weighted error calculation is performed on each detection unit to obtain detection unit evaluation data; Performing grading processing on the evaluation data of the detection unit to obtain quality grade distribution, and performing error grade visualization processing according to the quality grade distribution to obtain surface quality distribution data; Linear transition is performed on boundaries of adjacent detection units in the surface quality distribution data, an error coefficient of the transition area is calculated to obtain continuous distribution data, and a core surface quality distribution map is generated based on the continuous distribution data.

5. The method for controlling the processing accuracy of the injection mold core according to claim 4, characterized in that: The core surface quality distribution map is input into a dynamic compensation model including a local compensation amount calculation module and a compensation parameter optimization module to calculate the compensation coefficient of each region to obtain a compensation optimization parameter matrix, including: Constructing a plurality of compensation areas based on the core surface quality distribution map, and establishing a local compensation amount calculation module based on the plurality of compensation areas, calculating a pressure compensation coefficient, a speed compensation coefficient, and a time compensation coefficient of each compensation area, and obtaining a regional compensation coefficient matrix; According to the regional compensation coefficient matrix and the quality grade distribution, the grinding pressure deviation, the grinding speed deviation and the grinding time deviation of each compensation area are calculated to obtain a compensation parameter reference value; Establishing a dynamic compensation model based on the compensation parameter reference value, and calculating a first compensation parameter set through the dynamic compensation model; Performing parameter constraint judgment on the first compensation parameter set to obtain a second compensation parameter set, and performing weighted operation on the second compensation parameter set and the area weight coefficient of each compensation area to obtain a third compensation parameter set; The third compensation parameter set is iteratively optimized by using a dual particle swarm algorithm in a compensation parameter optimization module to obtain a target compensation parameter set, and the target compensation parameter set is reconstructed into a compensation optimization parameter matrix according to the multiple compensation areas.

6. The method for controlling the processing accuracy of the injection mold core according to claim 5, characterized in that: The method of iteratively optimizing the third compensation parameter set by using the dual particle swarm algorithm in the compensation parameter optimization module to obtain a target compensation parameter set, and reconstructing the target compensation parameter set into a compensation optimization parameter matrix according to the multiple compensation areas, comprises: Performing double particle swarm initialization processing on the third compensation parameter set by using the double particle swarm algorithm in the compensation parameter optimization module to construct a first population and a second population, each population comprising 40 particles, and each particle comprising parameters of three dimensions: pressure compensation ΔP, velocity compensation ΔV, and time compensation ΔT; The fitness of the first population and the second population is calculated respectively, and the fitness function is f(x)=w1×|ΔP|+w2×|ΔV|+w3×|ΔT|, wherein f(x) is the fitness function, w1, w2, and w3 are respectively the pressure compensation weight coefficient, the speed compensation weight coefficient, and the time compensation weight coefficient, and the values ​​are 0.4, 0.3, and 0.3, respectively, ΔP, ΔV, and ΔT are respectively the pressure compensation amount, the speed compensation amount, and the time compensation amount, and a first fitness evaluation result of the first population and a second fitness evaluation result of the second population are obtained; sorting the first population based on the first fitness evaluation result to obtain a first optimized initial value, and sorting the second population based on the second fitness evaluation result to obtain a second optimized initial value; Performing a velocity update calculation based on the first optimized initial value to obtain a new velocity of the first particle, and performing a velocity update calculation based on the second optimized initial value to obtain a new velocity of the second particle; Performing a position update calculation according to the new velocity of the first particle to obtain a new position of the first particle, and performing a position update calculation according to the new velocity of the second particle to obtain a new position of the second particle; Performing boundary processing on the first new particle position and the second new particle position respectively, limiting the parameters exceeding the compensation range to a reasonable range, and obtaining a first effective compensation parameter and a second effective compensation parameter; Performing a population crossover operation based on the first effective compensation parameter and the second effective compensation parameter, exchanging information between high-quality particles in the first population and the second population, and generating a target compensation parameter set by using an arithmetic crossover method; The target compensation parameter set is reorganized according to the compensation area to construct an n×m dimensional compensation parameter matrix, where n is the number of compensation areas and is equal to the number of the multiple compensation areas, and m is the compensation parameter dimension and is equal to 3, corresponding to the pressure compensation amount, the speed compensation amount and the time compensation amount, respectively, to obtain the compensation optimization parameter matrix.

7. The method for controlling the processing accuracy of the injection mold core according to claim 6, characterized in that: The multi-axis linkage path planning is performed based on the compensation optimization parameter matrix, the core surface is divided into a plurality of grid units, and the curvature value and the grinding trajectory compensation value of each grid unit are calculated respectively to generate a thermal deformation compensation grinding path, including: Based on the compensation optimization parameter matrix, the core surface is meshed to construct an 8×8 uniform mesh unit group, wherein the compensation parameters of each mesh unit include a pressure compensation amount ΔP, a velocity compensation amount ΔV, and a time compensation amount ΔT, and gridded compensation data is obtained; Calculating a local curvature value for each grid unit in the gridded compensation data, correcting a grinding pressure compensation amount according to the local curvature value to obtain a corrected pressure compensation amount, and adjusting a grinding speed compensation amount in a linkage manner based on the corrected pressure compensation amount to obtain a corrected speed compensation amount; Performing thermal effect analysis on the corrected pressure compensation amount and the corrected speed compensation amount, calculating the temperature rise value of each grid unit, obtaining thermal deformation compensation data, and correcting the grinding time compensation amount based on the thermal deformation compensation data to obtain a corrected time compensation amount; The corrected pressure compensation amount, the corrected speed compensation amount and the corrected time compensation amount are subjected to multi-axis linkage path planning, the motion trajectory of the grinding head in each grid unit is calculated to obtain a grinding path point sequence, and the grinding path point sequence is subjected to trajectory smoothing to obtain a thermal deformation compensation grinding path.

8. An injection mold core machining accuracy control system, characterized in that: Used to execute the method for controlling the machining accuracy of an injection mold core according to any one of claims 1 to 7, the injection mold core machining accuracy control system comprising: The acquisition module is used to acquire data on the core surface of the injection mold using a multi-point trigger sensor to obtain the core surface profile data, geometric dimension data and surface roughness data; A processing module, used for inputting the core surface profile data, the geometric dimension data and the surface roughness data into a feature extraction model, extracting the core surface fusion feature, and performing regional processing on the core surface to obtain regional grinding parameters; A calculation module, used for calculating a normalized error coefficient based on the regional grinding parameters and generating a core surface quality distribution map; An optimization module, used for inputting the core surface quality distribution map into a dynamic compensation model including a local compensation amount calculation module and a compensation parameter optimization module to calculate the compensation coefficient of each region, and obtain a compensation optimization parameter matrix; A generation module is used to perform multi-axis linkage path planning based on the compensation optimization parameter matrix, divide the core surface into multiple grid units, and calculate the curvature value and grinding trajectory compensation value of each grid unit respectively, and generate a thermal deformation compensation grinding path.

9. A computer device, characterized in that: The method comprises a memory and a processor, wherein the memory stores a computer program that can be run on the processor, and is characterized in that when the processor executes the computer program, the method for controlling the machining accuracy of the injection mold core according to any one of claims 1 to 7 is implemented. 10 . A computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the processor is enabled to execute the method for controlling the machining accuracy of an injection mold core according to any one of claims 1 to 7 .

Images