Auxiliary positioning method and system in multi-process switching of precision mold

Abstract

The application provides an auxiliary positioning method and system in multi-process switching of a precision mold, relates to the technical field of precision machining, and comprises the following steps: collecting process reference information of the mold in the current process and constructing a three-dimensional mold grid model; taking positioning reference data as a reference, performing grid comparison and offset calculation on the three-dimensional mold grid model to obtain real-time offset information; aiming at eliminating the real-time offset information, performing one-time position adjustment on the mold to obtain a one-time corrected mold; and performing two-time position adjustment on the one-time corrected mold according to historical positioning errors of the next adjacent process to execute mold machining of the next adjacent process. The application solves the technical problem that, due to the lack of real-time pose offset sensing and historical error compensation mechanism, the positioning error of the mold accumulates and the machining precision decreases in the prior art, significantly improves the positioning precision of the mold in multi-process switching, enhances the stability and reliability of multi-process switching, and improves the mold machining quality.

Description

Technical Field

[0001] This application relates to the field of precision machining technology, specifically to an auxiliary positioning method and system for multi-process switching of precision molds. Background Technology

[0002] In the field of precision manufacturing, mold processing typically involves the continuous switching of multiple processes, such as roughing, finishing, surface treatment, and assembly and debugging. To ensure that the mold maintains a consistent machining benchmark and clamping accuracy across different processes, multi-process switching positioning technology has become a key element in guaranteeing processing quality and efficiency.

[0003] Currently, positioning methods for multi-process switching of precision molds mostly rely on fixed reference blocks, mechanical locating pins, or tooling fixtures to achieve mold realignment manually or semi-automatically. While these methods are simple in structure and low in cost, they suffer from significant problems such as limited positioning accuracy, low operational efficiency, and the inability to detect and compensate for errors in a timely manner. Especially during mold switching, factors such as temperature changes, clamping stress, and the mold's own micro-deformation can easily cause slight shifts in mold position. Traditional methods struggle to detect these errors in a timely manner, leading to their gradual amplification in subsequent processes and affecting overall machining accuracy. Summary of the Invention

[0004] This application provides an auxiliary positioning method and system for multi-process switching of precision molds, which solves the technical problem that the lack of real-time positional offset perception and historical error compensation mechanism for molds during multi-process switching leads to the accumulation of mold positioning errors and a decrease in processing accuracy. It achieves the technical effect of significantly improving the positioning accuracy of molds during multi-process switching, enhancing the stability and reliability of multi-process switching, and improving the processing quality of molds.

[0005] In view of the above problems, this application provides an auxiliary positioning method for multi-process switching of precision molds. The method includes: during mold processing, collecting the process reference information of the mold in the current process and establishing positioning reference data; when the mold enters the process switching stage, using multi-source sensing devices to collect information about the mold and constructing a three-dimensional mold mesh model; using the positioning reference data as a reference, using the three-dimensional mold mesh model to perform mesh comparison and offset calculation to obtain real-time offset information; with the goal of eliminating the real-time offset information, controlling the auxiliary positioning device to perform a first position adjustment on the mold to obtain a first-corrected mold; based on the historical positioning error of the next adjacent process, using a multi-axis adjustment device to perform a second position adjustment on the first-corrected mold to complete the mold position correction and execute the mold processing of the next adjacent process.

[0006] On the other hand, this application also provides an auxiliary positioning system for multi-process switching of precision molds. The system includes: a positioning reference construction module, used to collect the process reference information of the mold in the current process and establish positioning reference data during mold processing; a mold information acquisition module, used to collect information of the mold using multi-source sensing devices and construct a three-dimensional mold mesh model when the mold enters the process switching stage; an offset calculation module, used to perform mesh comparison and offset calculation using the positioning reference data as a reference and the three-dimensional mold mesh model to obtain real-time offset information; a primary correction module, used to control the auxiliary positioning device to perform a primary position adjustment of the mold with the goal of eliminating the real-time offset information, to obtain a primary corrected mold; and a secondary correction module, used to perform a secondary position adjustment of the primary corrected mold using a multi-axis adjustment device based on the historical positioning error of the next adjacent process, to complete the mold position correction and execute the mold processing of the next adjacent process.

[0007] One or more technical solutions provided in this application have at least the following beneficial effects: By collecting the process reference information of the mold during the current process and establishing positioning reference data, an accurate reference standard is provided for positioning during subsequent process switching, ensuring the consistency and continuity of data between multiple processes. When the mold enters the process switching stage, multi-source sensing devices are used to collect information about the mold and construct a three-dimensional mold mesh model to realistically reflect the actual position and shape of the mold after the switch, providing a detailed data foundation for subsequent comparison and calculation. Based on the positioning reference data, the three-dimensional mold mesh model is used for mesh comparison and offset calculation to achieve quantitative analysis of the difference between the actual state of the mold and the theoretical benchmark, obtain accurate real-time offset information, and provide a decision basis for subsequent positioning correction. With the goal of eliminating the real-time offset information, the auxiliary positioning device is controlled to perform a first position adjustment on the mold, quickly correcting the main deviation direction of the mold and achieving preliminary accuracy recovery, resulting in a first-corrected mold. Based on the historical positioning error of the next adjacent process, the first-corrected mold is adjusted a second time through a multi-axis adjustment device to complete the mold position correction, improve the overall positioning stability and feedforward compensation capability, ensure the consistency of the starting point of the next process, and then execute the mold processing of the next adjacent process.

[0008] In summary, this application achieves high-precision assisted positioning of precision molds during multi-process switching by integrating a dual mechanism of real-time perception and historical data-driven approaches. The solution utilizes process reference data and a 3D mesh model to calculate the mold's pose offset in real time. Through a two-stage position adjustment strategy of current offset correction and historical error compensation, it significantly improves the positioning consistency and compensation accuracy during mold process transitions. This not only effectively avoids the error accumulation problem caused by insufficient perception capabilities of traditional methods but also enhances the adaptability to minute deformations and displacements under complex working conditions. Ultimately, it achieves the technical effect of significantly improving the positioning accuracy of molds during multi-process switching, enhancing the stability and reliability of multi-process switching, and improving the quality of mold processing.

[0009] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the following are specific embodiments of this application. Attached Figure Description

[0010] Figure 1 This is a flowchart illustrating the auxiliary positioning method for multi-process switching of precision molds provided in the embodiments of this application.

[0011] Figure 2 This is a schematic diagram illustrating the process of constructing a three-dimensional mold mesh model in the auxiliary positioning method for multi-process switching of precision molds provided in the embodiments of this application.

[0012] Figure 3 This is a schematic diagram of the auxiliary positioning system for multi-process switching of precision molds provided in the embodiments of this application.

[0013] Explanation of reference numerals in the attached drawings: Positioning reference construction module 10, mold information acquisition module 20, offset calculation module 30, primary correction module 40, secondary correction module 50. Detailed Implementation

[0014] This application provides an auxiliary positioning method and system for multi-process switching of precision molds, which solves the technical problem in the prior art that the lack of real-time positional offset perception and historical error compensation mechanism for molds during multi-process switching leads to the accumulation of mold positioning errors and a decrease in processing accuracy. It achieves the technical effect of significantly improving the positioning accuracy of molds during multi-process switching, enhancing the stability and reliability of multi-process switching, and improving the processing quality of molds.

[0015] Example 1, as Figure 1 As shown in the embodiment of this application, an auxiliary positioning method is provided in the multi-process switching of precision molds. The method includes: Step S100: During mold processing, collect the process reference information of the mold in the current process and establish positioning reference data.

[0016] Specifically, process reference information refers to the geometric elements and parameters upon which the mold's machining or clamping position is determined in the current process. This typically includes reference points, reference surfaces, reference lines, and process parameters such as machining and clamping. During the initial machining stage of the mold, a sensor system installed on the CNC equipment or inspection table collects key reference features of the mold in the current process in real time, including geometric information such as reference surfaces (e.g., bottom surface or clamping surface), reference lines (e.g., symmetry center line), and reference points (e.g., positioning hole center). Simultaneously, a machining model is established by combining machining parameters (e.g., tool path, spindle speed, feed rate) and clamping parameters (e.g., clamping force, clamping position). Based on this, a local coordinate system of the mold is constructed using the collected reference geometric information, and its posture is fine-tuned according to the process parameters. This generates positioning reference data with high-precision descriptive capabilities, enabling accurate capture and standardized expression of the mold's position and posture in the initial machining stage. This provides a high-quality reference basis for posture comparison and correction during subsequent multi-process switching, significantly improving the accuracy of subsequent positioning compensation. For example, in the first milling operation of an automotive mold, an XY plane datum is established using the centers of four positioning holes as reference points, and the positioning edge of the fixture is used as the Z-axis reference plane. The machining parameters are a feed rate of 300 mm / min, a spindle speed of 2000 rpm, and a clamping force of 1200 N. Finally, a positioning reference coordinate system containing a 6-dimensional pose description is established.

[0017] Step S200: When the mold enters the process switching stage, use multi-source sensing equipment to collect information about the mold and construct a three-dimensional mold mesh model.

[0018] Specifically, before the current process ends and the next process begins, multi-source sensing devices deployed around the workshop's measurement area or processing platform are used to collect information about the mold. These devices include image sensors, laser scanners, and structured light equipment. Image sensors capture images of the mold's surface texture and features from multiple angles; laser scanners acquire point cloud data through beam scanning, covering the entire shape of the mold. Contour extraction and feature point localization are performed on the image data, and the data is fused with the point cloud data to form a complete 3D mold surface model. Subsequently, the mesh size is set according to the process accuracy requirements, and the 3D surface is meshed to generate a structurally continuous 3D mesh model for subsequent comparison and analysis. Through multi-source sensing and high-precision 3D modeling, the true state of the mold during process switching can be fully restored, providing a digital foundation for subsequent pose offset analysis and dynamic correction, ensuring the comprehensiveness and accuracy of error identification.

[0019] Step S300: Using the positioning reference data as a benchmark, perform mesh comparison and offset calculation using the three-dimensional mold mesh model to obtain real-time offset information.

[0020] Furthermore, the real-time offset information includes coordinate position offset and attitude angle deviation.

[0021] Specifically, the positioning reference data established in step S100 is used as the target template, and the three-dimensional mold mesh model obtained in step S200 is used as the current state. The two are compared point-to-point using the ICP (Iterative Closest Point) algorithm. The comparison process calculates the spatial offset of each mesh point and summarizes it into real-time offset information including coordinate position offset and attitude angle deviation. Among them, the coordinate position offset includes the displacement change in the three-dimensional space (X, Y, Z directions) of the mold, and the attitude angle deviation includes the change in the rotation attitude (pitch angle, yaw angle, roll angle) of the mold, which can be integrated into six-degree-of-freedom offset parameters: three position offsets (ΔX, ΔY, ΔZ) and three attitude angle deviations (ΔPitch, ΔYaw, ΔRoll), where ΔX is the displacement deviation in the X-axis direction, ΔY is the displacement deviation in the Y-axis direction, ΔZ is the displacement deviation in the Z-axis direction, ΔPitch is the pitch angle deviation, ΔYaw is the yaw angle deviation, and ΔRoll is the roll angle deviation. This step enables high-precision detection and parameterized representation of the current pose error of the mold, providing a clear quantitative target for subsequent automated position adjustment and significantly improving the error response speed and compensation accuracy during process switching.

[0022] Step S400: With the goal of eliminating the real-time offset information, control the auxiliary positioning device to adjust the position of the mold once to obtain a corrected mold.

[0023] Specifically, the auxiliary positioning device refers to a mechanical or electrically controlled structure used to precisely adjust the mold position, such as a pneumatic positioning pin, a six-degree-of-freedom flexible positioning platform, or a servo control mechanism. After receiving the offset information output in step S300, the adjustment path is calculated based on the mold's mass and structural characteristics, and the auxiliary positioning device is driven to correct the position. The auxiliary positioning device performs small displacements or attitude adjustments on the mold according to the set path, ensuring the mold enters the error tolerance range (e.g., within ±0.05mm, ±0.1°). After adjustment, the current mold state is marked as "first-time correction mold" for subsequent, more refined historical error correction, reducing the complexity of subsequent fine-tuning and improving mold switching efficiency and the stability of the correction effect.

[0024] Step S500: Based on the historical positioning error of the next adjacent process, the mold position is adjusted twice by the multi-axis adjustment device to complete the mold position correction, and the mold processing of the next adjacent process is executed.

[0025] Specifically, historical positioning error is high-frequency offset information statistically obtained from positioning adjustment records of the same or similar molds in similar processes in the past. A multi-axis adjustment device is a positioning platform or robot with multiple degrees of freedom adjustment capabilities, enabling high-precision spatial pose adjustment.

[0026] To further improve positioning accuracy, after one correction, historical positioning error data of similar molds in the adjacent process stored on the industrial internet platform are consulted, and an error prediction model is established based on the current geometric deviation of the mold. According to the offset trend output by this model, small-range, high-precision fine-tuning (e.g., ±0.01mm) is performed using a multi-axis adjustment device, ultimately bringing the mold posture close to the theoretical optimal state. After adjustment, the mold is locked in the new position and enters the next processing step. This step introduces historical data feedback and a high-precision multi-axis adjustment mechanism, making the positioning strategy adaptive and forward-looking, effectively preventing error accumulation, and improving the continuity of process connections and processing accuracy.

[0027] Furthermore, step S100 includes: Step S110: Collect the reference points, reference surfaces, reference lines and process parameters of the mold in the current process as the process reference information, wherein the process parameters include machining parameters and clamping parameters.

[0028] Step S120: Establish the reference coordinate system and target attitude angle of the mold based on the reference point, reference surface and reference line.

[0029] Step S130: Fine-tune the reference coordinate system and target attitude angle according to the processing parameters and clamping parameters to obtain the positioning reference data.

[0030] Specifically, in the initial mold-making process, geometric datum elements such as reference points, reference surfaces, and reference lines of the mold, as well as process parameters such as machining parameters and clamping parameters, are collected by connecting to the machine tool control system, sensor interface, or process design system (such as CAM software) for this process. Machining parameters include parameters required for CNC machining such as feed rate, spindle speed, and cutting path, while clamping parameters include process conditions for fixing the mold such as clamping force, clamping method, and clamping position. For example, vision sensors and probes are used to identify the center of the positioning hole and the clamping plane; the controller reads the machining speed and tool path from the G-code; and the fixture monitoring system records the clamping force and contact point position in real time. All this information constitutes the complete process datum information for this process.

[0031] Based on the collected reference geometric elements, a logical coordinate system is determined in the mold model. For example, a reference point is set as the origin, a reference plane defines the Z-axis direction, a reference line defines the X-axis direction, and the Y-axis is automatically derived to complete the right-handed construction of the coordinate system. Simultaneously, by comparing the model's design attitude, the ideal target attitude angles required for the current process (such as 0° yaw, 0° roll, -3° pitch) are generated. This transforms discrete geometric features into a standardized coordinate system and attitude definition, enabling a quantifiable expression of the mold state and providing a stable reference for offset detection and automatic correction.

[0032] Due to machining errors, clamping deformation, or mold loading deviations, there may be slight offsets between the original reference coordinate system and the attitude angle. Therefore, simulation calculations are performed by combining machining parameters (such as spindle thrust direction and tool contact area) and clamping parameters (force direction at clamping points and fixture stiffness) to generate a fine-tuning vector. This vector is then applied to the original reference coordinate system to dynamically correct the reference coordinate system and the target attitude angle, obtaining positioning reference data. This achieves an adaptive correction mechanism that combines ideal attitude with actual working conditions, improving the fit and usability of the reference data, thereby enhancing the accuracy and robustness of subsequent mold position recognition and correction actions.

[0033] Furthermore, such as Figure 2 As shown, step S200 includes: Step S210: After the current process is completed, use an image sensor to acquire images of the mold at a preset acquisition angle to obtain a mold image set, and use a laser scanner to scan the mold in all directions to obtain a mold point cloud set.

[0034] Step S220: Extract contour lines based on the mold image set and fuse them to construct a three-dimensional mold contour.

[0035] Step S230: Use the three-dimensional mold contour to perform point cloud fitting on the mold point cloud set to construct a three-dimensional mold model.

[0036] Step S240: Divide the three-dimensional mold model into a mesh according to the preset mesh size to obtain the three-dimensional mold mesh model.

[0037] Specifically, after the current process is completed, the image acquisition and scanning process is initiated. First, under the control system, the image sensor surrounds the mold at a preset angle (e.g., rotating every 30°) to acquire images and generate a multi-view image set. Then, the laser scanner performs high-precision laser ranging on the entire surface of the mold to obtain the original point cloud of the mold, providing a foundation for subsequent 3D model construction.

[0038] Image processing algorithms (such as Canny edge detection, Sobel operator, and deep edge recognition model) are used to analyze the mold image set and extract contour feature lines. These contour lines are then converted into three-dimensional coordinates using camera calibration data (such as stereo vision reconstruction and beam triangulation), thus forming the preliminary three-dimensional contour of the mold. These contours can be used as structural boundary references in subsequent fusion processes with point cloud data.

[0039] The obtained 3D contour is used as a structural constraint. The obtained point cloud is then fitted with a surface and reconstructed topologically to convert the mold point cloud into a continuous surface model. A 3D mold model is then constructed to realize the transformation of the mold surface from discrete points to a continuous surface, ensuring that the model has good visualization effects and applicability for subsequent analysis, and improving the expression accuracy of the mold model.

[0040] Based on preset mesh size parameters, a mesh generation algorithm (such as Delaunay triangulation, octree partitioning, etc.) is used to mesh the 3D surface, generating a 3D mold mesh model with equal or adaptive density. The preset mesh size is determined according to the actual processing conditions. This model features both lightweight structure and high precision, facilitating subsequent offset analysis, attitude comparison, or simulation verification. Through meshing, the 3D mold model obtains a standardized structural form, enabling efficient participation in subsequent comparison, error analysis, and attitude adjustment calculations, significantly improving system efficiency and analysis accuracy.

[0041] Furthermore, step S220 includes: Step S221: Using a set of sample mold images and a set of sample mold contours of the same type of mold, train a convolutional neural network until convergence to obtain a contour line extractor.

[0042] Step S222: Using the contour line extractor, extract contour lines from multiple mold images in the mold image set to obtain multiple mold contours.

[0043] Step S223: Based on the preset acquisition angle, perform feature registration and smoothing repair according to the multiple mold contours to construct a three-dimensional mold contour.

[0044] Specifically, the sample mold image set consists of pre-collected multi-angle image data of known molds of the same type. The sample mold contour set is a set of mold edge lines manually annotated or automatically extracted, serving as training labels. Using supervised learning methods, a large number of images of similar molds and their corresponding contour labels are used as training data to train a convolutional neural network model to learn the edge structure features in the images, resulting in a contour line extractor. An example of the training process is as follows: First, sample mold images of similar molds and their corresponding contour line annotation images are collected and organized to construct training sample pairs. Each sample includes mold images from multiple preset collection angles and their corresponding contour line annotations. The annotation process uses image annotation software to draw pixel-level contour edges and uniformly converts them into binary mask images. To enhance the model's generalization ability, data augmentation operations, including rotation, flipping, brightness perturbation, blurring, and affine transformation, are performed on the image samples to expand the sample quantity and visual diversity. Subsequently, a convolutional neural network model with U-Net as the backbone is constructed, with RGB mold images as input and contour probability maps as output. The downsampling part of the network uses ResNet-18 as the feature extraction skeleton, while the upsampling part uses a multi-layer deconvolution module for resolution restoration. During training, a weighted combination of binary cross-entropy loss (BCE), Dice loss, and edge structure loss is used as the total loss function to enhance the model's accuracy in recognizing contour edges. The model is trained using the Adam optimizer with an initial learning rate of 0.001. Eight samples are selected per training run, and the training epochs are 100. The training and validation sets are divided in an 8:2 ratio. Model performance is monitored using metrics such as intersection-over-union ratio (IoU), F1 score, and edge offset error. Once the model's performance on the validation set converges and stabilizes, the final trained model weight file is saved as a contour line extractor for subsequent batch contour extraction tasks on mold images.

[0045] The set of mold images acquired by the image sensor is input into the contour line extractor. Edge detection, filtering, and contour localization operations are performed on each image, and a contour map matching the image size is output. The set of contour maps corresponding to all images constitutes multiple mold contours, preparing for subsequent 3D modeling.

[0046] Using the preset acquisition angle during image capture, 3D reconstruction projections of contour lines from different angles are performed. Subsequently, ICP (Iterative Closest Point) registration algorithms and feature point matching are employed to spatially align the contours from each viewpoint. Curvature filtering and smoothing interpolation are performed on edge lines with defects, burrs, or breaks to achieve contour repair. Finally, these are combined into a continuous and complete 3D contour framework, achieving a unified representation of the mold's multi-angle image contours in 3D space. This generates a structurally continuous and precision-controlled 3D contour framework, providing high-quality geometric guidance information for subsequent point cloud fitting.

[0047] Furthermore, step S230 includes: Step S231: Using the three-dimensional mold contour, perform a random fitting on the mold point cloud set to obtain a first point cloud fitting result, and calculate the ratio of the number of point clouds falling into the first point cloud fitting result to the total number of point clouds in the mold point cloud set to obtain the first fitting accuracy.

[0048] Step S232: Using the three-dimensional mold contour again, perform a second random fitting on the mold point cloud to obtain a second point cloud fitting result and a second fitting accuracy.

[0049] Step S233: Perform iterative fitting until the preset number of fittings is reached, output the point cloud fitting result with the maximum fitting accuracy as the optimal point cloud fitting result, and construct a three-dimensional mold model based on the optimal point cloud fitting result.

[0050] Specifically, before fitting the mold point cloud set, a pre-constructed 3D mold contour is used as a reference model, and the Random Sample Consensus Algorithm (RANSAC) is employed for preliminary point cloud fitting. The specific steps are as follows: several points in the mold point cloud are randomly selected as initial seed points for fitting, and these seed points are used to generate a preliminary planar or curved surface fitting model. Then, the distance between each point in the mold point cloud and the fitted model is calculated, and it is determined whether it meets a preset tolerance threshold. Points that meet the condition are considered to fall into the fitting result, and the number of these points is recorded. The ratio of the number of point clouds falling into the fitting result to the total number of point clouds in the mold point cloud set is calculated to obtain the accuracy of the first fitting. The higher this ratio, the better the fitting effect and the higher the accuracy.

[0051] After the initial fitting, to further improve the fitting accuracy, the point cloud is fitted again using the 3D mold contour. This time, the same Random Sample Consensus (RANSAC) algorithm as the first time is used, but the initial seed points and tolerance values ​​are adjusted to further refine the fitting results. The new fitting model generates new point cloud fitting results, and the fitting accuracy is recalculated. Similarly, the standard for calculating fitting accuracy is the ratio of the number of points falling into the fitted result to the total number of points in the point cloud. If the accuracy of the second fitting result is higher than that of the first, then the point cloud fitting result generated at this time is the second fitting result.

[0052] The above steps are followed to iterate and fit the data until the preset maximum number of fitting iterations is reached. The fitting accuracy is calculated for each point cloud fitting result. When the preset limit is reached, the fitting result with the highest fitting accuracy is selected as the optimal point cloud fitting result. Finally, a 3D mold model is constructed based on this optimal point cloud fitting result, ensuring that the geometry of this model matches the actual mold's point cloud data to the greatest extent possible, providing a high-quality model foundation for subsequent mesh generation and positioning analysis.

[0053] Furthermore, the method for generating the preset mesh size in step S240 includes: Step S240-1: During the current processing step, monitor and acquire the processing parameter sequence and clamping parameter sequence.

[0054] Step S240-2: Using the standard machining parameters and standard clamping parameters of the current process as a benchmark, perform deviation analysis on the machining parameter sequence and clamping parameter sequence to obtain the machining parameter deviation amplitude sequence and clamping parameter deviation amplitude sequence.

[0055] Step S240-3: Perform mean calculation and fluctuation analysis on the processing parameter deviation amplitude sequence, and output the mean of processing parameter deviation amplitude and the processing parameter fluctuation coefficient, wherein the fluctuation coefficient is the ratio of the standard deviation to the mean.

[0056] Step S240-4: Calculate the mean and analyze the fluctuation of the clamping parameter deviation amplitude sequence, and output the mean of the clamping parameter deviation amplitude and the clamping parameter fluctuation coefficient.

[0057] Step S240-5: The current machining deviation scale is calculated by weighting the average value of the machining parameter deviation amplitude, the machining parameter fluctuation coefficient, the average value of the clamping parameter deviation amplitude, and the clamping parameter fluctuation coefficient.

[0058] Step S240-6: Calculate the ratio of the preset processing deviation scale of the current process to the current processing deviation scale, multiply it by the initial grid size, and obtain the preset grid size.

[0059] Specifically, during the current machining process, various machining and clamping parameters are continuously monitored and collected. These parameters include cutting speed, feed rate, tool position, workpiece position, and clamping force. These parameter sequences are recorded in real time through sensors, control systems, or data acquisition devices, providing necessary raw data for subsequent deviation analysis.

[0060] After obtaining the machining parameter sequence and clamping parameter sequence for the current process, a deviation analysis is performed on the actual machining parameter sequence and clamping parameter sequence, using the standard machining parameters and standard clamping parameters for the current process as a benchmark. Specifically, the deviation magnitude of each machining parameter and clamping parameter relative to the standard value is calculated, and a machining parameter deviation magnitude sequence and a clamping parameter deviation magnitude sequence are generated. These deviation magnitude data can be used to reflect the instability or error sources in the machining process.

[0061] Next, the mean and fluctuation analysis are performed on the processing parameter deviation amplitude sequence. First, the mean of the processing parameter deviation is obtained, which represents the average degree of deviation. Then, the fluctuation coefficient is derived by calculating the standard deviation of the deviation sequence. The fluctuation coefficient formula is: Fluctuation coefficient = Standard deviation / Mean. The fluctuation coefficient describes the fluctuation amplitude of the processing parameters, that is, the degree of parameter change during actual processing. The smaller the mean of the processing parameter deviation amplitude, the lower the fluctuation coefficient, indicating a more stable processing process and smaller error.

[0062] Similarly, the mean and fluctuation analysis are performed on the clamping parameter deviation amplitude sequence. The mean of the clamping parameter deviation is obtained, and its fluctuation coefficient is calculated. The deviation analysis of the clamping parameters can reveal possible errors and instabilities in the clamping process. Likewise, the smaller the mean deviation amplitude and the smaller the fluctuation coefficient, the higher the stability of the clamping process.

[0063] The machining deviation scale for the current process is obtained by weighting the mean deviation amplitude of machining parameters, the fluctuation coefficient of machining parameters, the mean deviation amplitude of clamping parameters, and the fluctuation coefficient of clamping parameters. This machining deviation scale integrates the influence of all deviation amplitudes and fluctuation coefficients, and can accurately reflect the overall situation of actual machining errors in the process. Generally speaking, the smaller the deviation, the smaller the fluctuation, and the smaller the machining deviation scale.

[0064] Finally, the ratio of the preset machining deviation to the actual machining deviation is calculated. This ratio reflects the deviation between the machining accuracy of the current process and the preset standard. Multiplying this ratio by the initially set mesh size yields the final preset mesh size. If the deviation is large, the mesh size needs to be reduced to improve the mesh division accuracy and ensure higher machining precision. Conversely, if the deviation is small, the mesh size can be appropriately increased to optimize computational efficiency. The initial mesh size is set based on the mold size and initial accuracy requirements, serving as the benchmark for adjusting the preset mesh size.

[0065] Through the above steps, the preset grid size can be dynamically adjusted according to the actual deviation of the processing parameters and clamping parameters, ensuring that the mold positioning and error correction during the processing can achieve the required accuracy and stability.

[0066] Furthermore, step S500 includes: Step S510: Use the three-dimensional mold mesh model to perform mesh comparison and offset calculation, obtain the geometric shape deviation of the mold under the current process, and expand the geometric shape deviation according to the preset deviation tolerance range to obtain the geometric shape deviation range.

[0067] Step S520: Using the geometric deviation range and the next adjacent process as retrieval constraints, retrieve similar mold processing information based on the Industrial Internet to determine the sample historical positioning error set.

[0068] Step S530: Based on the analysis of the historical positioning error set of the sample, determine the historical positioning error, and with the goal of eliminating the historical positioning error, perform a secondary position adjustment on the primary correction mold through a multi-axis adjustment device.

[0069] Specifically, the preset deviation tolerance range is an allowable deviation range set according to process requirements and experience. It is used to expand the geometric deviation, forming a geometric deviation range. Using the currently generated 3D mold mesh model, mesh comparison and offset calculations are performed. By comparing the actual geometry of the mold in the current process with the preset geometry (usually a standard model or reference model), the geometric deviation of the mold is calculated, including dimensional deviations, angular deviations, etc. Based on the preset deviation tolerance range, the geometric deviation is expanded to form a geometric deviation range. This expansion is to cover possible error ranges, ensuring that even if the deviation fluctuates, it can still be tolerated and compensated accordingly. In this way, it can be ensured that the mold can still be maintained within a suitable tolerance range during secondary adjustments, avoiding other problems caused by over-correction.

[0070] Using the obtained geometric deviation range as a retrieval constraint, and combined with the processing requirements of the next adjacent process, the industrial internet platform is used to retrieve processing information for similar molds, querying processing data for similar molds under similar processes. From these retrieved historical data, a sample historical positioning error set is selected. This error set contains positioning error information experienced by other molds in the same or similar processes. By analyzing these historical positioning errors, reference data can be provided for the secondary adjustment of the current mold.

[0071] Statistical analysis is performed on the historical positioning error data from the sample to determine the historical positioning errors. Based on this, a correction scheme aimed at eliminating historical positioning errors is formulated, and secondary position adjustments are performed. A multi-axis adjustment device is used to perform reverse adjustment of the mold (i.e., advance compensation). By pre-setting the compensation amount before the process, the impending error is eliminated, thereby reducing the accumulation of errors and effectively improving the machining accuracy of the next process.

[0072] Furthermore, step S530 includes: Step S531: Obtain the sample historical positioning error set, wherein the sample historical positioning error includes X-axis offset data, Y-axis offset data, Z-axis offset data, pitch angle deviation, yaw angle deviation and roll angle deviation.

[0073] Step S532: Randomly select the first sample historical positioning error. If the proportion of any deviation value in the first sample historical positioning error that is less than a preset deviation threshold is greater than a preset proportion, then the first sample historical positioning error is set as a high-frequency historical positioning error. The high-frequency historical positioning error set is obtained by sequentially filtering.

[0074] Step S533: Calculate the mean of the high-frequency historical positioning error set to obtain the historical positioning error.

[0075] Specifically, positioning data of similar molds in the same or similar processes are retrieved to obtain a sample historical positioning error set. Each sample's historical positioning error data includes X-axis offset data, Y-axis offset data, Z-axis offset data, pitch angle deviation, yaw angle deviation, and roll angle deviation. Among them, X-axis offset data is the positional offset of the mold in the X-axis direction; Y-axis offset data is the positional offset of the mold in the Y-axis direction; Z-axis offset data is the positional offset of the mold in the Z-axis direction; pitch angle deviation is the tilt angle deviation of the mold around the X-axis direction; yaw angle deviation is the rotation angle deviation of the mold around the Z-axis direction; and roll angle deviation is the rotation angle deviation of the mold around the Y-axis direction.

[0076] A first sample of historical positioning error is randomly selected from the historical positioning error set. Then, the deviation value of this first sample is compared with the deviation magnitudes of other similar data. If, in this sample, the proportion of deviations in any direction less than a preset deviation threshold is greater than a preset percentage (e.g., 80% of the deviations are less than the set threshold), then this error sample is considered a high-frequency historical positioning error, meaning it is a common positioning error. In practice, the deviation value of the first X-axis offset data of the first sample historical positioning error is first calculated compared with the deviation values ​​of other X-axis offset data in the historical positioning error set. The proportion of data less than a preset deviation threshold is counted. If this proportion is greater than a preset percentage, then the proportions of Y-axis offset data, Z-axis offset data, pitch angle deviation, yaw angle deviation, and roll angle deviation are calculated sequentially. If all these proportions are greater than preset percentages, then the first sample historical positioning error is set as a high-frequency historical positioning error. This selection process is repeated to select a series of high-frequency historical positioning error sets from the historical positioning error set. These errors will serve as reference data for the correction target.

[0077] The mean value of the selected high-frequency historical positioning error set is calculated to obtain a representative historical positioning error value. This historical positioning error value can reflect the average level of positioning error in the same mold processing process. It provides an accurate target value for the secondary position adjustment of the first-corrected mold through the multi-axis adjustment device, which helps to eliminate historical positioning errors more accurately and improve the precision of mold processing.

[0078] In summary, the auxiliary positioning method for multi-process switching of precision molds provided in this application has the following beneficial effects: This application's embodiments achieve high-precision assisted positioning of precision molds during multi-process switching by integrating a dual mechanism of real-time perception and historical data-driven approaches. Real-time offset calculation of the mold's pose is achieved using process reference data and a 3D mesh model. A two-stage position adjustment strategy—current offset correction and historical error compensation—significantly improves positioning consistency and compensation accuracy during mold process transitions. This not only effectively avoids the error accumulation problem caused by insufficient perception capabilities of traditional methods but also enhances the adaptability to minute deformations and displacements under complex working conditions. Ultimately, this achieves the technical effect of significantly improving the positioning accuracy of molds during multi-process switching, enhancing the stability and reliability of multi-process switching, and improving mold processing quality.

[0079] Example 2, as Figure 3 As shown, based on the same inventive concept as in Embodiment 1 above, this application provides an auxiliary positioning system for multi-process switching of precision molds, the system comprising: The positioning reference construction module 10 is used to collect the process reference information of the mold in the current process and establish positioning reference data during mold processing.

[0080] The mold information acquisition module 20 is used to collect information about the mold using multi-source sensing devices when the mold enters the process switching stage, and to construct a three-dimensional mold mesh model.

[0081] The offset calculation module 30 is used to perform mesh comparison and offset calculation using the positioning reference data as a reference and the three-dimensional mold mesh model to obtain real-time offset information.

[0082] Furthermore, the real-time offset information includes coordinate position offset and attitude angle deviation.

[0083] The primary correction module 40 is used to control the auxiliary positioning device to perform a primary position adjustment on the mold with the goal of eliminating the real-time offset information, thereby obtaining a primary corrected mold.

[0084] The secondary correction module 50 is used to adjust the position of the primary correction mold a second time through a multi-axis adjustment device based on the historical positioning error of the next adjacent process, so as to complete the mold position correction and execute the mold processing of the next adjacent process.

[0085] Furthermore, in this embodiment of the application, the positioning reference construction module 10 is also used to perform the following steps: The reference points, reference surfaces, reference lines, and process parameters of the mold in the current process are collected as the process reference information, wherein the process parameters include machining parameters and clamping parameters; a reference coordinate system and target attitude angle of the mold are established based on the reference points, reference surfaces, and reference lines; the reference coordinate system and target attitude angle are fine-tuned according to the machining parameters and clamping parameters to obtain the positioning reference data.

[0086] Furthermore, in this embodiment of the application, the mold information acquisition module 20 is also used to perform the following steps: After the current process is completed, an image sensor is used to acquire images of the mold at a preset acquisition angle to obtain a mold image set. The mold is then scanned from all directions by a laser scanner to obtain a mold point cloud set. Contour lines are extracted from the mold image set and fused to construct a three-dimensional mold contour. The mold point cloud set is then fitted with the three-dimensional mold contour to construct a three-dimensional mold model. The three-dimensional mold model is then divided into a mesh according to a preset mesh size to obtain the three-dimensional mold mesh model.

[0087] Furthermore, in this embodiment of the application, the mold information acquisition module 20 is also used to perform the following steps: A convolutional neural network is trained to convergence using a sample mold image set and a sample mold contour set of similar molds to obtain a contour line extractor. The contour line extractor is used to extract contour lines from multiple mold images in the mold image set to obtain multiple mold contours. Based on the preset acquisition angle, feature registration and smoothing repair are performed on the multiple mold contours to construct a three-dimensional mold contour.

[0088] Furthermore, in this embodiment of the application, the mold information acquisition module 20 is also used to perform the following steps: Using the three-dimensional mold contour, a first random fitting is performed on the mold point cloud set to obtain a first point cloud fitting result. The ratio of the number of points falling into the first point cloud fitting result to the total number of points in the mold point cloud set is calculated to obtain a first fitting accuracy. The three-dimensional mold contour is then used again to perform a second random fitting on the mold point cloud set to obtain a second point cloud fitting result and a second fitting accuracy. Iterative fitting is performed until a preset number of fittings is reached. The point cloud fitting result with the maximum fitting accuracy is output as the optimal point cloud fitting result, and a three-dimensional mold model is constructed based on the optimal point cloud fitting result.

[0089] Furthermore, the system described in this application embodiment is also used to perform the following steps: During the current processing step, the processing parameter sequence and clamping parameter sequence are monitored and acquired. Using the standard processing parameters and standard clamping parameters of the current process as a benchmark, deviation analysis is performed on the processing parameter sequence and clamping parameter sequence to obtain the processing parameter deviation amplitude sequence and clamping parameter deviation amplitude sequence. The processing parameter deviation amplitude sequence is then subjected to mean calculation and fluctuation analysis, outputting the mean processing parameter deviation amplitude and processing parameter fluctuation coefficient, where the fluctuation coefficient is the ratio of the standard deviation to the mean. The clamping parameter deviation amplitude sequence is also subjected to mean calculation and fluctuation analysis, outputting the mean clamping parameter deviation amplitude and clamping parameter fluctuation coefficient. The current processing deviation scale is calculated by weighting the mean processing parameter deviation amplitude, processing parameter fluctuation coefficient, mean clamping parameter deviation amplitude, and clamping parameter fluctuation coefficient. The ratio of the preset processing deviation scale of the current process to the current processing deviation scale is calculated and multiplied by the initial grid size to obtain the preset grid size.

[0090] Furthermore, in this embodiment of the application, the secondary correction module 50 is also used to perform the following steps: The three-dimensional mold mesh model is used for mesh comparison and offset calculation to obtain the geometric shape deviation of the mold in the current process. The geometric shape deviation is then expanded according to a preset deviation tolerance range to obtain a geometric shape deviation range. Using the geometric shape deviation range and the next adjacent process as search constraints, information on similar mold processing is retrieved based on the Industrial Internet to determine the sample historical positioning error set. The historical positioning error is determined based on the sample historical positioning error set. With the goal of eliminating the historical positioning error, the mold is repositioned a second time using a multi-axis adjustment device.

[0091] Furthermore, in this embodiment of the application, the secondary correction module 50 is also used to perform the following steps: A sample historical positioning error set is obtained, wherein the sample historical positioning error includes X-axis offset data, Y-axis offset data, Z-axis offset data, pitch angle deviation, yaw angle deviation, and roll angle deviation; a first sample historical positioning error is randomly selected. If the proportion of any deviation value in the first sample historical positioning error that is less than a preset deviation threshold compared with other similar data in the sample historical positioning error set is greater than a preset proportion, then the first sample historical positioning error is set as a high-frequency historical positioning error. The high-frequency historical positioning error set is obtained by sequentially filtering; the mean of the high-frequency historical positioning error set is calculated to obtain the historical positioning error.

[0092] Through the foregoing detailed description of the auxiliary positioning method in the multi-process switching of precision molds, those skilled in the art can clearly understand that the auxiliary positioning system in the multi-process switching of precision molds in this embodiment has corresponding functional modules and beneficial effects as it corresponds to the method disclosed in Embodiment 1. For relevant details, please refer to the description in the method section.

[0093] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. An auxiliary positioning method for multi-process switching in precision molds, characterized in that, The methods include: During mold processing, the process reference information of the mold in the current process is collected, and positioning reference data is established; When the mold enters the process changeover stage, multi-source sensing devices are used to collect information about the mold and construct a three-dimensional mold mesh model. Based on the positioning reference data, the three-dimensional mold mesh model is used to perform mesh comparison and offset calculation to obtain real-time offset information; With the goal of eliminating the real-time offset information, the auxiliary positioning device is controlled to adjust the position of the mold once to obtain a corrected mold; Based on the historical positioning error of the next adjacent process, the mold position is adjusted twice by the multi-axis adjustment device to complete the mold position correction, and the mold processing of the next adjacent process is executed.

2. The auxiliary positioning method for multi-process switching of precision molds according to claim 1, characterized in that, Collect the process reference information of the mold in the current process and establish positioning reference data, including: Collect the reference points, reference surfaces, reference lines, and process parameters of the mold in the current process as the process reference information, wherein the process parameters include machining parameters and clamping parameters; Establish the reference coordinate system and target attitude angle of the mold based on the reference points, reference surfaces and reference lines; The reference coordinate system and target attitude angle are finely adjusted based on the processing parameters and clamping parameters to obtain the positioning reference data.

3. The auxiliary positioning method for multi-process switching of precision molds according to claim 1, characterized in that, Information is collected from the mold using multi-source sensing devices to construct a three-dimensional mold mesh model, including: After the current process is completed, the mold image set is obtained by using an image sensor according to the preset acquisition angle. The mold is then scanned in all directions by a laser scanner to obtain the mold point cloud set. Based on the mold image set, contour lines are extracted and fused to construct a three-dimensional mold contour. The point cloud of the mold is fitted using the three-dimensional mold contour to construct a three-dimensional mold model. The three-dimensional mold model is divided into meshes according to a preset mesh size to obtain the three-dimensional mold mesh model.

4. The auxiliary positioning method for multi-process switching of precision molds according to claim 3, characterized in that, Based on the mold image set, contour lines are extracted and fused to construct a three-dimensional mold contour, including: Using a set of sample mold images and a set of sample mold contours of the same type of mold, a convolutional neural network is trained until convergence to obtain a contour line extractor. Using the contour line extractor, contour lines are extracted from multiple mold images in the mold image set to obtain multiple mold contours; Based on the preset acquisition angle, feature registration and smoothing repair are performed on the multiple mold contours to construct a three-dimensional mold contour.

5. The auxiliary positioning method for multi-process switching of precision molds according to claim 3, characterized in that, Using the three-dimensional mold contour, a point cloud fitting is performed on the mold point cloud set to construct a three-dimensional mold model, including: Using the three-dimensional mold contour, a random fitting is performed on the mold point cloud set to obtain a first point cloud fitting result, and the ratio of the number of point clouds falling into the first point cloud fitting result to the total number of point clouds in the mold point cloud set is calculated to obtain the first fitting accuracy. The three-dimensional mold contour is used again to perform a second random fitting on the mold point cloud to obtain a second point cloud fitting result and a second fitting accuracy. Perform iterative fitting until the preset number of fittings is reached, output the point cloud fitting result with the maximum fitting accuracy as the optimal point cloud fitting result, and construct a three-dimensional mold model based on the optimal point cloud fitting result.

6. The auxiliary positioning method for multi-process switching of precision molds according to claim 3, characterized in that, The method for generating the preset grid size includes: During the current processing step, the processing parameter sequence and clamping parameter sequence are monitored and acquired. Based on the standard machining parameters and standard clamping parameters of the current process, a deviation analysis is performed on the machining parameter sequence and clamping parameter sequence to obtain the machining parameter deviation amplitude sequence and the clamping parameter deviation amplitude sequence. The mean and fluctuation analysis are performed on the processing parameter deviation amplitude sequence, and the mean of the processing parameter deviation amplitude and the processing parameter fluctuation coefficient are output, wherein the fluctuation coefficient is the ratio of the standard deviation to the mean. The mean value and fluctuation analysis of the clamping parameter deviation amplitude sequence are performed to output the mean value of the clamping parameter deviation amplitude and the clamping parameter fluctuation coefficient. The current machining deviation magnitude is calculated by weighting the average deviation magnitude of the machining parameters, the fluctuation coefficient of the machining parameters, the average deviation magnitude of the clamping parameters, and the fluctuation coefficient of the clamping parameters. Calculate the ratio of the preset processing deviation scale of the current process to the current processing deviation scale, multiply it by the initial grid size, and obtain the preset grid size.

7. The auxiliary positioning method for multi-process switching of precision molds according to claim 1, characterized in that, The real-time offset information includes coordinate position offset and attitude angle deviation.

8. The auxiliary positioning method for multi-process switching of precision molds according to claim 1, characterized in that, Based on the historical positioning error of the next adjacent process, the primary correction mold is repositioned a second time using a multi-axis adjustment device, including: The three-dimensional mold mesh model is used to perform mesh comparison and offset calculation to obtain the geometric shape deviation of the mold under the current process, and the geometric shape deviation is expanded according to the preset deviation tolerance range to obtain the geometric shape deviation range. Using the geometric deviation range and the next adjacent process as retrieval constraints, information retrieval of similar mold processing is performed based on the Industrial Internet to determine the sample historical positioning error set; Based on the analysis of the historical positioning error set of the sample, the historical positioning error is determined, and with the goal of eliminating the historical positioning error, the position of the primary correction mold is adjusted a second time through a multi-axis adjustment device.

9. The auxiliary positioning method for multi-process switching of precision molds according to claim 8, characterized in that, Historical positioning errors are determined based on the analysis of the historical positioning error set of the sample, including: Obtain the sample historical positioning error set, which includes X-axis offset data, Y-axis offset data, Z-axis offset data, pitch angle deviation, yaw angle deviation, and roll angle deviation; Randomly select the first sample historical positioning error. If the proportion of any deviation value in the first sample historical positioning error that is less than a preset deviation threshold with respect to other similar data in the sample historical positioning error set is greater than a preset proportion, then the first sample historical positioning error is set as a high-frequency historical positioning error. The high-frequency historical positioning error set is obtained by sequentially filtering. The historical positioning error is obtained by averaging the high-frequency historical positioning error set.

10. An auxiliary positioning system for multi-process switching in precision molds, characterized in that, The system is used to execute the auxiliary positioning method in multi-process switching of precision molds as described in any one of claims 1-9, including: The positioning reference construction module is used to collect the process reference information of the mold in the current process and establish positioning reference data during mold processing. The mold information acquisition module is used to collect information about the mold using multi-source sensing devices when the mold enters the process switching stage, and to build a three-dimensional mold mesh model. The offset calculation module is used to perform mesh comparison and offset calculation using the positioning reference data as a benchmark, and to obtain real-time offset information using the three-dimensional mold mesh model. A primary correction module is used to control the auxiliary positioning device to perform a primary position adjustment on the mold with the goal of eliminating the real-time offset information, thereby obtaining a primary corrected mold. The secondary correction module is used to adjust the position of the primary correction mold a second time through a multi-axis adjustment device based on the historical positioning error of the next adjacent process, so as to complete the mold position correction and execute the mold processing of the next adjacent process.

Images