A large precision part machining precision self-adaptive detection method and related equipment

Abstract

The present application relates to a kind of large precision parts machining precision self-adaptive detection method and related equipment.The method includes: the part to be tested is divided into multiple detection regions and determines each region importance index;Reference piece is set on the non-machining reference surface of part or tooling;Initial detection is executed to obtain initial detection result and state data;Based on state data, construct multidimensional evaluation index, calculate measurement reliability coefficient;According to importance index and reliability coefficient, determine the priority of retest, divide the region into direct acquisition area, encryption re-scan area and contact recheck area;Additional detection is executed to the area needing retest;According to deviation information, the detection result is corrected at regional level;Finally, the modified result is differentially fused and output.The method dynamically evaluates the measurement reliability by dividing the region, realizes the data quality intensive guarantee of key area, and improves the overall detection precision and efficiency of large parts.

Description

Technical Field

[0001] This application relates to the field of precision measurement and quality inspection technology, and in particular to an adaptive detection method and related equipment for the machining accuracy of large precision parts. Background Technology

[0002] Large gears and other precision rotating parts are core transmission components of high-end equipment. The machining accuracy of their tooth profile, tooth direction, tooth pitch, and reference surface directly determines the transmission smoothness, load-bearing capacity, and service life of the entire mechanical system. With the advancement of flexible manufacturing and digital production lines, the inspection of large precision parts is gradually evolving from independent offline sampling inspection to embedded full inspection on the production line. This places extremely high demands on the measurement accuracy of the inspection process and the production line cycle time.

[0003] In some existing technologies, the machining accuracy inspection of large gears employs coordinate measuring machines (CMMs) for contact probing or non-contact sensors such as lidar for contour scanning. In practice, these inspection devices primarily rely on fixed measurement paths preset offline based on the workpiece's 3D CAD model to perform scanning or touch-testing actions. To cope with fluctuations in the workshop environment, periodic manual calibration or global error compensation methods based on static polynomial fitting are often used to offset systematic errors caused by environmental parameters such as temperature and vibration.

[0004] However, in practical applications, the surface morphology of large gears is complex, and different regions (such as the core meshing tooth surface, mounting reference surface, and non-working end face) have significantly different levels of importance in actual physical function and varying sensitivities to tolerances. Furthermore, in real-world workshop production environments, changes in temperature gradients, minute vibrations of the testing platform, and slight fluctuations in the workpiece clamping state occur in real time and dynamically. These environmental and equipment disturbances do not uniformly interfere with the measurement of different local areas of the large workpiece, resulting in differences in the reliability of measurement results for different local areas during the same testing process. Summary of the Invention

[0005] This application provides an adaptive detection method and related equipment for machining accuracy of large precision parts, which can improve the accuracy of machining accuracy detection of large precision parts.

[0006] Firstly, this application provides an adaptive detection method for the machining accuracy of a large precision part under test. The method includes: acquiring model information and detection requirements of the large precision part under test, and dividing the large precision part under test into multiple detection areas; determining the importance index of each detection area based on the model information and detection requirements; setting reference parts or extracting reference features on the non-machining reference surface of the large precision part under test or on its clamping fixture; performing initial detection on the large precision part under test to obtain initial detection results, and collecting state data, which includes deviation information of the reference parts or reference features relative to the nominal state; extracting environmental disturbance and clamping stability features based on the state data, and combining the data quality features of the initial detection results to construct a multi-dimensional... The evaluation sub-indicators are normalized to determine the measurement reliability coefficient of each detection area. Based on the importance index and the measurement reliability coefficient, the retest priority of each detection area is determined, and each detection area is divided into at least two categories: direct acceptance area, local encrypted rescan area, and contact verification area. Additional detection is performed on the local encrypted rescan area and / or contact verification area to obtain additional detection results. According to the deviation information, the initial detection results and additional detection results are corrected at the regional level to obtain corrected initial detection results and corrected additional detection results. Replacement processing or weighted fusion processing is performed on the corrected initial detection results and corrected additional detection results to output the final detection result of the large precision part to be tested.

[0007] This embodiment extracts environmental and clamping features from state data and constructs evaluation sub-indicators based on data quality features to determine the measurement reliability coefficient of each detection area. Then, it classifies retest priorities and area categories based on importance indicators. This transforms a single global processing approach into a differentiated measurement strategy based on local real-state conditions. Additional detection and area-level correction are specifically performed on low-reliability or high-importance areas, avoiding resource waste or insufficient sampling in critical areas caused by full-scale retesting. This improves the accuracy of overall detection data for large parts and the efficiency of detection resource allocation.

[0008] In some embodiments of the first aspect, the method involves extracting environmental disturbance and clamping stability features based on state data, and combining this with the data quality features of the initial detection results to construct multi-dimensional evaluation sub-indicators. After normalizing these sub-indicators, the measurement reliability coefficient for each detection area is determined. Specifically, this includes: parsing state data, extracting the temperature gradient change rate and platform vibration amplitude, and calculating the environmental disturbance degree sub-indicator; extracting the displacement of the clamping point of the large precision part before and after detection, and calculating the clamping stability degree sub-indicator; comparing the deviation information with historical benchmark deviations, and calculating the reference consistency degree sub-indicator; calculating the initial detection consistency degree sub-indicator based on the point cloud noise distribution and local curvature variance of each detection area in the initial detection results; assigning preset weights to the sub-indicators, and normalizing the weighted summation results to obtain the measurement reliability coefficient for each detection area. This measurement reliability coefficient is negatively correlated with the temperature gradient change rate, platform vibration amplitude, displacement, and point cloud noise distribution.

[0009] This embodiment constructs sub-indices for environmental disturbance, clamping stability, reference consistency, and initial detection consistency by analyzing the temperature gradient change rate and platform vibration amplitude, extracting clamping point displacement, comparing historical benchmark deviations, and analyzing point cloud noise distribution. This multi-dimensional quantitative evaluation method integrates macroscopic environmental factors, equipment physical state, and microscopic data characteristics, enabling the final determined measurement reliability coefficient to accurately reflect the comprehensive impact of multi-source interference on local measurement results, thus improving the objectivity and comprehensiveness of regional reliability assessment.

[0010] In conjunction with some embodiments of the first aspect, in some embodiments, the extraction of the displacement of the clamping point of the large precision part to be tested before and after testing, and the calculation of the clamping stability sub-index, specifically includes: acquiring the clamping point coordinate sequence in real time during the initial testing process using an optical target or displacement sensor arranged at the clamping fixture; calculating the maximum offset of the coordinate sequence as the displacement; if the maximum offset exceeds a preset elastic deformation threshold, determining that clamping loosening has occurred and identifying the target clamping point where clamping loosening has occurred; constructing a spatial penalty function centered on the target clamping point and exhibiting distance decay for the target clamping point; and using the spatial penalty function to penalize and amplify the clamping stability sub-index of the detection area adjacent to the target clamping point, so as to reduce the measurement reliability coefficient of the corresponding detection area.

[0011] This embodiment acquires the coordinate sequence of clamping points in real time and extracts target clamping points that exceed the elastic deformation threshold, constructing a spatial penalty function with distance decay centered on these points. This spatial penalty function is used to penalize and amplify the clamping stability sub-index of the adjacent detection area, transforming local physical loosening events into a quantitative weakening of the measurement reliability of the surrounding area. This allows the system to adaptively identify and isolate local data severely affected by clamping deformation, reducing the risk of measurement distortion in the surrounding area due to local clamp failure.

[0012] In conjunction with some embodiments of the first aspect, in some embodiments, the retesting priority of each detection area is determined based on the importance index and the measurement reliability coefficient, and each detection area is divided into at least two categories: direct acceptance area, local encrypted rescanning area, and contact verification area. Specifically, this includes: multiplying the importance index of each detection area by the reciprocal of the measurement reliability coefficient to obtain the comprehensive retesting priority score of each detection area as the retesting priority; setting a first priority threshold and a second priority threshold, wherein the first priority threshold is less than the second priority threshold; if the comprehensive retesting priority score of a detection area is less than the first priority threshold, it is classified as a direct acceptance area; if the comprehensive retesting priority score of a detection area is greater than or equal to the first priority threshold and less than the second priority threshold, it is classified as a local encrypted rescanning area; if the comprehensive retesting priority score of a detection area is greater than or equal to the second priority threshold, it is classified as a contact verification area.

[0013] This embodiment obtains a comprehensive retest priority score by multiplying the importance index of the detection area by the reciprocal of the measurement reliability coefficient, and sets first and second priority thresholds to divide the area into intervals. This calculation logic mathematically amplifies the weight of high-importance, low-reliability areas, enabling the system to automatically allocate areas to direct acceptance areas, locally encrypted rescan areas, or contact verification areas based on the score intervals. This establishes a decision-making mechanism that balances functional tolerance requirements and real-time measurement quality, optimizing the scheduling logic of retest tasks.

[0014] In conjunction with some embodiments of the first aspect, in some embodiments, the initial detection result and the additional detection result are corrected at the regional level based on the deviation information to obtain the corrected initial detection result and the corrected additional detection result. Specifically, this includes: constructing a global spatial error compensation matrix based on the deviation information; calculating the spatial distance from the centroid of each detection region to the reference part or reference feature for each detection region, and generating a regional attenuation factor based on the spatial distance; decomposing the global spatial error compensation matrix into translation error components and rotation error components; using the regional attenuation factor to perform nonlinear scaling processing on the translation error components and rotation error components respectively to suppress the leverage effect of rotation error amplification with spatial distance, to obtain local translation error components and local rotation error components; reconstructing the local transformation matrix for each detection region based on the local translation error components and local rotation error components; and using the local transformation matrix to perform rigid body transformation on the initial detection result and the additional detection result respectively to obtain the corrected initial detection result and the corrected additional detection result.

[0015] This embodiment generates a region-level attenuation factor by calculating the spatial distance from the centroid of the detection area to the reference part. After decomposing the global spatial error compensation matrix, the attenuation factor is used to nonlinearly scale the translational and rotational error components respectively. This mechanism dynamically adjusts the error compensation intensity based on the spatial correlation between each region and the reference, suppressing the unbounded amplification of rotational errors in the far-end region. The reconstructed local transformation matrix makes the spatial error compensation more consistent with the physical deformation law of large parts, improving the rationality of region-level error correction.

[0016] In conjunction with some embodiments of the first aspect, in some embodiments, the replacement processing or weighted fusion processing is performed on the corrected initial detection result and the corrected additional detection result to output the final detection result of the large precision part to be tested. Specifically, this includes: for the direct acceptance area, directly outputting the corrected initial detection result as the final detection result; for the local densified rescanning area, registering the corrected additional detection result with the corrected initial detection result in point cloud, and performing weighted average fusion based on the sampling point density distribution of the two to obtain the final detection result; for the contact verification area, using the contact measurement point of the corrected additional detection result as a reference, performing forced constraint deformation and resampling on the local point cloud of the corresponding corrected initial detection result to complete the data replacement; at the boundary edge of the direct acceptance area, the local densified rescanning area and the contact verification area, constructing confidence weights based on the measurement reliability coefficients of the sampling points in adjacent areas, and using a weighted distance inverse interpolation algorithm to perform gridded stitching on the data in the transition area to output the final detection result.

[0017] This embodiment performs differentiated processing based on the characteristics of different regions. In the locally encrypted rescanning area, weighted fusion is performed based on point density. In the contact verification area, forced constraint deformation is performed based on the contact measurement points. At the boundary edge, the measurement reliability coefficients of sampling points in adjacent areas are extracted to construct confidence weights. A weighted inverse distance interpolation algorithm is used for grid stitching. This joint interpolation mechanism based on reliability and spatial distance smooths the transition boundary between heterogeneous data sources, stitching the locally differentiated measurement results into a three-dimensional topographic model with global continuity, ensuring the topological integrity of the final detection result.

[0018] In conjunction with some embodiments of the first aspect, in some embodiments, before obtaining the model information and testing requirements of the large precision part to be tested and dividing the large precision part to be tested into multiple testing areas, the method further includes: performing topological structure and mass distribution analysis on the model information of the large precision part to be tested; identifying multiple rigid nodes in the large precision part to be tested; setting the location of the rigid node as the mounting position of the reference part, or extracting the inherent geometric features at the rigid node as reference features.

[0019] Before dividing the inspection area, this embodiment analyzes the topology and mass distribution of the large precision part under test using model information. This identifies the rigid node with the smallest deformation under external forces, and the reference mounting position or reference feature extraction point is set at this node. This preliminary planning mechanism, based on the physical structural characteristics of the part itself, ensures that the established nominal state benchmark has structural disturbance resistance throughout the entire inspection cycle, providing a stable spatial coordinate system reference for subsequent state data acquisition and spatial error compensation.

[0020] In a second aspect, embodiments of this application provide an adaptive detection device for the machining accuracy of large precision parts. The adaptive detection device for the machining accuracy of large precision parts includes: one or more processors and a memory; the memory is coupled to the one or more processors, and the memory is used to store computer program code, the computer program code including computer instructions, and the one or more processors call the computer instructions to cause the adaptive detection device for the machining accuracy of large precision parts to perform the method described in the first aspect and any possible implementation thereof.

[0021] Thirdly, embodiments of this application provide a computer program product containing instructions that, when the computer program product is run on a large precision parts machining accuracy adaptive detection device, cause the large precision parts machining accuracy adaptive detection device to execute the method described in the first aspect and any possible implementation thereof.

[0022] Fourthly, embodiments of this application provide a computer-readable storage medium including instructions that, when executed on a large precision parts machining accuracy adaptive detection device, cause the large precision parts machining accuracy adaptive detection device to perform the method described in the first aspect and any possible implementation thereof.

[0023] Understandably, the adaptive detection equipment for machining accuracy of large precision parts provided in the second aspect, the computer program product provided in the third aspect, and the computer storage medium provided in the fourth aspect are all used to execute the methods provided in the embodiments of this application. Therefore, the beneficial effects they can achieve can be referred to the beneficial effects in the corresponding methods, and will not be repeated here.

[0024] One or more technical solutions provided in the embodiments of this application have at least the following technical effects or advantages:

[0025] 1. This application determines the measurement reliability of each area by combining status data and data quality characteristics, and divides the retest priority by combining importance indicators. It transforms the single global processing into a differentiated measurement strategy, avoids the waste of resources caused by full retesting, and improves the accuracy of the overall detection data and the efficiency of resource allocation.

[0026] 2. This application constructs evaluation sub-indicators by analyzing multi-dimensional state data such as temperature gradient, platform vibration and clamping point displacement, and integrates macro-environment, equipment status and micro-data for calculation, so that the measurement reliability coefficient can truly reflect the comprehensive impact of multi-source dynamic interference on the local measurement results of a large workpiece.

[0027] 3. This application nonlinearly scales the translation and rotation components of the global error compensation matrix by generating a regional attenuation factor, and dynamically adjusts the compensation intensity based on spatial correlation, thereby suppressing the unbounded amplification of rotation error in the far region and making spatial error compensation more in line with the physical deformation law of large parts. Attached Figure Description

[0028] Figure 1 This is a flowchart illustrating an adaptive detection method for machining accuracy of large precision parts in an embodiment of this application.

[0029] Figure 2 This is a schematic diagram of the physical device structure of an adaptive detection device for machining accuracy of large precision parts, as described in this application. Detailed Implementation

[0030] The terminology used in the following embodiments of this application is for the purpose of describing particular embodiments only and is not intended to be limiting of this application. As used in the specification of this application, the singular expressions “a,” “an,” “the,” “the,” and “this” are intended to include the plural expressions as well, unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used in this application refers to any or all possible combinations including one or more of the listed items.

[0031] Hereinafter, the terms "first" and "second" are used for descriptive purposes only and should not be construed as implying or suggesting relative importance or implicitly indicating the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature, and in the description of the embodiments of this application, unless otherwise stated, "multiple" means two or more.

[0032] To facilitate understanding, the method provided in this implementation will be described in detail below, using the above scenario as an example. Please refer to [link / reference]. Figure 1 This is a flowchart illustrating an adaptive detection method for machining accuracy of large precision parts in this application.

[0033] S101. Obtain the model information and testing requirements of the large precision part to be tested, and divide the large precision part to be tested into multiple testing areas.

[0034] The model information refers to the digital geometric representation data of the large precision part to be tested, including a 3D CAD solid model, theoretical tooth profile and direction parameters, datum coordinate system definition, tolerance zone annotation information, and key functional surface identification data. The inspection requirements specify the measurement task specifications to be performed on the part, including at least a list of inspection items, allowable deviation ranges for each item, sampling point density and accuracy requirements, and inspection cycle time constraints. The inspection area represents several spatial sub-regions formed by dividing the surface of the part to be tested according to functional attributes or geometric features; each inspection area serves as the basic unit for subsequent measurement resource allocation and data processing.

[0035] After acquiring model information and inspection requirements, the system needs to discretize the continuous part surface into a finite number of inspection areas to implement a differentiated inspection strategy for each region. Specifically, region division can adopt a hierarchical division method based on functional surfaces and tolerance levels, setting the tooth meshing area, tooth root transition area, end face reference area, and mounting hole group area as independent inspection areas, and further subdividing them according to the tolerance level of each area; or a curvature analysis method based on geometric topology can be used, using the feature boundaries with drastic local curvature changes as the partition boundaries. For large gear parts, each tooth groove can be divided into an independent area and marked with a tooth number; for toothed flanges or slewing bearings, functional surfaces such as mounting stops and sealing grooves need to be extracted separately as high-precision inspection areas. This division method allows the system to allocate differentiated measurement resources to areas of different functional importance, avoiding resource waste or insufficient sampling in key areas caused by using a uniform sampling density on the entire surface.

[0036] In some preferred embodiments, prior to the above steps, the method further includes: performing topological and mass distribution analysis on the model information of the large precision part to be tested; identifying multiple rigid nodes in the large precision part to be tested; setting the location of the rigid node as the mounting position of the reference part, or extracting the inherent geometric features at the rigid node as reference features.

[0037] Topology analysis refers to the process of calculating and analyzing the geometric connection relationships, facet adjacency relationships, and voxel connectivity of the 3D CAD model of the part under test. Mass distribution analysis represents the process of calculating the degree of mass concentration and the position of the centroid in various local areas of the part based on the material density and geometric volume of the part. Rigid nodes refer to feature points or regions in the part structure with significantly higher local stiffness than the surrounding areas and minimal deformation under external forces. These nodes are usually located in thick-walled connection areas, reinforcing rib intersection areas, near datum hole groups, or mass concentration areas. The mounting position of the reference part indicates the mounting position of an independent reference part on the non-machined surface or in the corresponding area of ​​the clamping fixture at the location of the rigid node. Inherent geometric features represent the stable geometric elements that already exist at the rigid node, including features such as the center of the datum hole, the axis of the locating pin hole, the unmachined original surface, or the blank datum surface retained before machining.

[0038] Before acquiring model information and testing requirements, the system pre-identifies the most suitable rigid node locations for arranging reference datums. Specifically, it performs topological analysis on the 3D CAD model of the large precision part to be tested, extracts the facet adjacency matrix and voxel connectivity graph, and identifies regions with high structural closure. It then performs mass distribution analysis on the model, calculates the mass moments of each local region based on material density and volume, and identifies areas of concentrated mass. Combining the topological and mass distribution information, it employs a finite element static analysis method to apply simulated clamping forces and measure contact forces to the part, calculating the displacement response of each node. Nodes with displacement responses less than 0.01 mm are marked as candidate rigid nodes. From these candidate rigid nodes, nodes located within 30 mm of thick-walled connection areas, reinforcing rib intersection areas, or reference hole groups are selected, ultimately identifying 3 to 5 rigid nodes.

[0039] For identified rigid nodes, if the node is located in a flat area such as the thick-walled area of ​​the gear spokes or the intersection area of ​​flange reinforcing ribs, this location is set as the mounting position for a reference part, and a ball target or coded target is pasted at this location as an independent reference part. If the node is located in an area with existing geometric features such as the gear center hole or the locating pin hole, the center coordinates of the reference hole or the axis of the locating pin hole are directly extracted as the reference feature, without the need to install additional reference parts.

[0040] S102. Determine the importance indicators of each detection area based on model information and detection requirements.

[0041] The importance index is used to represent the relative importance of each inspection area in the overall functional realization of the part. This index is a regional-level quantified value and can be ranked and compared. This index comprehensively considers multiple dimensions of factors such as functional attributes, tolerance level, force transmission path, meshing contact state, and assembly datum relationship, reflecting the influence weight of the machining accuracy deviation of the area on the overall performance of the part.

[0042] Once the model information and testing requirements are clear, the system needs to calculate the importance index for each testing area to establish a basis for subsequent resource allocation decisions. Specifically, a weighting method based on functional attributes can be adopted. According to the functional hierarchy of the meshing area, transmission contact surface, assembly reference surface, and non-contact area, weight coefficients of 0.9, 0.8, 0.7, and 0.3 are assigned respectively, and then adjusted in conjunction with the tolerance level of the area. The higher the tolerance level, the greater the weight is multiplied by an amplification factor of 1.2 to 1.5.

[0043] Another approach is a risk weighting method based on historical data. This method extracts the frequency of defects and the severity of failures in each region of historical batches, increasing the importance index of high-frequency defect areas or severe failure areas by 20% to 40%. Taking large gears as an example, the importance index of the working area of ​​the gear tooth surface, which is directly involved in power transmission and typically has a tolerance grade of IT5 to IT6, can be set to 0.85 to 0.95; while the importance index of the gear spokes or non-working end faces can be set to 0.2 to 0.4.

[0044] S103. Set a reference or extract a reference feature on the non-machined reference surface of the large precision part to be tested or on its clamping fixture.

[0045] The reference part refers to an independent physical standard part mounted on the unmachined surface of the part under test or on the clamping fixture, including ball targets, coded targets, standard blocks, or reference pins, etc. This reference part has known and stable geometric features. Reference features are used to represent stable geometric elements directly extracted from the body of the part under test, including inherent features such as unmachined original surfaces, reference hole centers, locating pin hole axes, or blank reference surfaces retained before machining. The nominal state refers to the theoretical pose relationship of the reference part or reference feature relative to the workpiece coordinate system under an ideal, undisturbed environment. This state is established through a high-precision calibration program before the start of the test.

[0046] Before performing the initial inspection, the system needs to place reference datums in the non-functional area of ​​the part under test or on its clamping fixture to monitor environmental disturbances and clamping changes in real time during the inspection process. Specifically, if an independent reference part scheme is used, three ball targets can be pasted on the non-machined area of ​​the outer circle of the gear, or an array of coded targets can be installed on the fixed end face of the clamping chuck. The reference part material should be an alloy material with a coefficient of thermal expansion close to that of the part under test. The installation method should use magnetic adsorption or threaded locking structure to prevent it from falling off during the inspection. The installation position should avoid the scanning field of view and maintain a safe distance of 30 to 50 mm from the main inspection area.

[0047] If a reference feature extraction scheme is adopted, the unmachined end face of the gear's inner hole or the center of the locating pin hole can be selected as the reference feature. Before inspection, the reference feature is measured three times repeatedly using a contact probe or a high-precision optical sensor to establish a nominal state reference. The establishment of the nominal state needs to be performed after the ambient temperature has stabilized. The measurement results are correlated with the workpiece coordinate system through a rigid body transformation matrix, which serves as the theoretical reference for deviation calculation in subsequent correction steps.

[0048] S104. Perform initial testing on the large precision parts to be tested to obtain initial test results and collect status data.

[0049] The initial inspection refers to the first round of measurement and scanning actions performed by the system on each inspection area of ​​the part under test. This inspection uses non-contact LiDAR or structured light sensors to collect contour point clouds of the entire surface or key areas. The initial inspection results represent the raw point cloud dataset acquired during the initial inspection, including the spatial coordinates, normal vectors, reflection intensity, and timestamp information of each inspection area. The status data represents the environmental and equipment status monitoring information collected synchronously with the initial inspection process, including the workshop temperature gradient change rate recorded by the temperature sensor, the vibration amplitude of the inspection platform captured by the vibration sensor, the fixture coordinate sequence output by the clamping point displacement sensor, and the deviation information of the current pose of the reference part or reference feature relative to the nominal state.

[0050] After completing the division of the detection area and setting the reference benchmark, the system starts the initial detection process, and performs measurement operations on the part under test according to the preset scanning path. Specifically, the initial detection uses a non-contact scanning device to collect point cloud data sequentially from important areas to secondary areas. The scanning step distance is adaptively adjusted according to the importance index of each area, with an important area using a step distance of 0.1 to 0.2 mm and a secondary area using a step distance of 0.5 to 1.0 mm.

[0051] During the scanning process, the system uses a time synchronization protocol with the sensor network to collect temperature data from multiple points in the workshop every 5 seconds from the temperature sensor and calculate the gradient change rate. It also collects platform acceleration signals recorded by the vibration sensor at a frequency of 1000Hz and extracts the peak vibration amplitude. The system acquires the coordinate sequence of the optical target at the clamping point to monitor whether the fixture has shifted. Simultaneously, it periodically remeasures the reference part or reference feature to obtain its real-time deviation relative to its nominal state. For areas deep in the gear tooth grooves or in areas of mirror reflection, the system automatically adjusts the laser incident angle or activates an auxiliary light source to avoid data loss.

[0052] S105. Based on the state data, extract the environmental disturbance and clamping stability features, and combine them with the data quality features of the initial detection results to construct multi-dimensional evaluation sub-indicators. After normalizing the evaluation sub-indicators, determine the measurement reliability coefficient of each detection area.

[0053] Among them, environmental disturbance characteristics are used to represent the degree of interference of temperature gradient and platform vibration on measurement stability during the detection process, and are quantitatively characterized by the rate of change of temperature gradient and the peak value of vibration amplitude. Clamping stability characteristics represent the displacement and deformation state of the clamping system of the part under test during the detection process, and are quantitatively evaluated by the maximum offset of the clamping point coordinate sequence. Data quality characteristics refer to the measurement consistency and noise level presented by the initial detection results themselves, and are quantitatively described by the standard deviation of point cloud noise distribution and the variance of local curvature. Evaluation sub-indices are used to represent several quantitative indicators that characterize the measurement reliability of the area from different dimensions, including environmental disturbance degree sub-indices, clamping stability degree sub-indices, reference consistency degree sub-indices, and initial detection consistency degree sub-indices. The measurement reliability coefficient represents the degree of credibility of the current measurement results in each detection area in truly reflecting the actual processing state of the part. This coefficient is a normalized value between 0 and 1, and the higher the value, the more reliable the measurement results in that area.

[0054] After acquiring the initial detection results and status data, the system needs to quantitatively evaluate the measurement reliability of each detection area to identify which areas are significantly affected by environmental and equipment conditions. Specifically, the system first analyzes the status data, extracts the maximum value of the temperature gradient change rate and the peak value of the platform vibration amplitude within the detection cycle, calculates the environmental disturbance degree sub-index as a weighted sum of the two and normalizes it to the 0-1 range, with the weighting coefficient determined based on the sensitivity coefficient during equipment calibration. Then, it extracts the clamping point coordinate sequence, calculates the maximum displacement of each clamping point before and after detection, and uses the maximum value of all clamping point displacements as the input for the clamping stability sub-index, which is negatively correlated with the displacement.

[0055] The system compares the real-time deviation information of the reference part or reference feature with the baseline deviation of historical batches, and calculates the deviation similarity as a sub-index of reference consistency. If the current deviation pattern is consistent with the historical pattern, the sub-index value is higher. Simultaneously, the system analyzes the point cloud noise distribution of each detection area in the initial detection results, and calculates the standard deviation of point cloud coordinates and the variance of local curvature. The smaller the value, the higher the scanning quality of the area, which serves as a sub-index of initial detection consistency. The system assigns preset weights to the above four sub-indexes. The weight values ​​can be determined through offline calibration experiments or optimized based on historical batch statistics. The weighted summation result is normalized using a Sigmoid function or a piecewise linear function to obtain the measurement reliability coefficient of each detection area.

[0056] In some preferred embodiments, the above steps specifically include: parsing state data, extracting temperature gradient change rate and platform vibration amplitude, and calculating environmental disturbance degree sub-index; extracting the displacement of the clamping point of the large precision part to be tested before and after testing, and calculating clamping stability sub-index; comparing the deviation information with historical benchmark deviation, and calculating reference consistency sub-index.

[0057] Based on the point cloud noise distribution and local curvature variance of each detection area in the initial detection results, the initial detection consistency sub-index is calculated; preset weights are assigned to the sub-index, and the weighted summation result is normalized to obtain the measurement reliability coefficient of each detection area. This measurement reliability coefficient is negatively correlated with the temperature gradient change rate, platform vibration amplitude, displacement, and point cloud noise distribution.

[0058] The environmental disturbance sub-index represents the degree of interference between temperature gradient and platform vibration on measurement stability during the testing process. It is quantified by the rate of change of temperature gradient and the peak value of vibration amplitude. The rate of change of temperature gradient represents the rate of change of temperature difference at multiple points in the workshop per unit time, measured in degrees Celsius per minute, and is obtained by collecting and calculating data from multi-channel temperature sensors arranged around the workpiece. The platform vibration amplitude represents the peak acceleration of the testing platform during the measurement process, measured in millimeters per second squared, and is acquired by a vibration sensor fixed to the base of the testing platform at a sampling frequency of 1000 Hz. The clamping stability sub-index represents the displacement and deformation state of the clamping system of the part under test, quantified by the maximum offset of the clamping point coordinate sequence, measured in millimeters. The reference consistency sub-index is obtained by calculating the cosine similarity between the current reference part deviation vector and the historical batch reference deviation vector, with a value between 0 and 1. The initial detection consistency sub-index is calculated based on the normalized reciprocal of the standard deviation of the point cloud coordinates and the local curvature variance of each detection area.

[0059] After acquiring the initial test results and status data, the system needs to quantitatively evaluate the measurement reliability of each test area. Specifically, the system analyzes the status data, extracts the maximum value of the temperature gradient change rate and the peak value of the platform vibration amplitude within the test cycle, and calculates the environmental disturbance degree sub-index as a weighted sum of 0.4 times the normalized value of the temperature gradient change rate and 0.6 times the normalized value of the vibration amplitude. The system extracts the clamping point coordinate sequence, calculates the maximum displacement of each clamping point in three-dimensional space before and after the test, and substitutes the maximum displacement of all clamping points into a negative exponential function to obtain the clamping stability sub-index. The system calculates the cosine similarity between the current deviation vector of the reference part and the historical batch deviation vector to obtain the reference consistency sub-index. Finally, the system analyzes the point cloud data of each test area in the initial test results, calculates the weighted sum of the standard deviation of the point cloud coordinates and the local curvature variance, and normalizes it by taking the reciprocal to obtain the initial test consistency sub-index.

[0060] The four sub-indicators were assigned preset weights of 0.25, 0.3, 0.2, and 0.25, respectively. The weighted summation result was mapped to the 0-1 interval using the Sigmoid function to obtain the measurement reliability coefficient for each detection area. This coefficient is negatively correlated with the temperature gradient change rate, platform vibration amplitude, clamping point displacement, and point cloud noise distribution; that is, the larger the values ​​of the above interference factors, the lower the measurement reliability coefficient.

[0061] In other embodiments of this application, the above-mentioned extraction of the displacement of the clamping point of the large precision part to be tested before and after testing, and the calculation of the clamping stability sub-index, specifically includes:

[0062] By using optical targets or displacement sensors placed at the clamping fixture, the coordinate sequence of the clamping point during the initial detection process is collected in real time; the maximum offset of the coordinate sequence is calculated as the displacement; if the maximum offset exceeds the preset elastic deformation threshold, it is determined that the clamping has become loose and the target clamping point where the clamping has become loose is identified.

[0063] For the target clamping point where clamping loosening occurs, a spatial penalty function with distance decay centered on the target clamping point is constructed. The spatial penalty function is used to penalize and amplify the clamping stability sub-index of the detection area adjacent to the target clamping point, so as to reduce the measurement reliability coefficient of the corresponding detection area.

[0064] The optical target refers to reflective markers or coded patterns placed on the clamping fixture, whose spatial position is captured in real time by a high-speed camera. The displacement sensor refers to a non-contact laser displacement sensor or inductive displacement sensor placed on the clamping jaws or support plate, outputting the three-dimensional coordinates of the clamping point in real time. The clamping point coordinate sequence represents the set of spatial coordinates of the clamping points recorded chronologically throughout the initial inspection process, with a sampling frequency of 10 to 20 times per second. The elastic deformation threshold represents the maximum allowable elastic deformation of the clamping system under normal clamping force; this threshold is set based on the fixture material, workpiece weight, and historical batch statistics, with a typical value of 0.02 to 0.05 mm. The spatial penalty function represents a weighting function centered on the target clamping point, decaying exponentially or Gaussianly with spatial distance, used to impose a reliability penalty on adjacent inspection areas.

[0065] When extracting the displacement of the clamping points, the system needs to determine whether the clamping system has become abnormally loose and to take targeted measures for the affected areas. Specifically, by using optical targets or displacement sensors placed at the clamping fixture, the system collects the coordinate sequence of the clamping points in real time at a frequency of 15 times per second during the initial detection process. After performing median filtering on the coordinate sequence to remove occasional noise, the maximum offset of each clamping point is calculated. If the maximum offset exceeds the preset elastic deformation threshold of 0.03 mm, the clamping point is determined to be loose and marked as the target clamping point.

[0066] For the target clamping point, a Gaussian attenuation spatial penalty function with a radius attenuation constant of 100 mm is constructed, centered at this point. The function value is 1.0 at the target clamping point, decays to 0.37 at a distance of 100 mm from the target clamping point, and approaches zero at a distance of 300 mm. This spatial penalty function is used to calculate the spatial distance between the target clamping point and the centroid of each detection area. Substituting the distance into the penalty function yields the penalty coefficient. The clamping stability sub-index of the detection area is multiplied by 1.5 to 2.0 times the penalty coefficient to amplify the penalty and reduce the measurement reliability coefficient of the corresponding detection area.

[0067] S106. Based on the importance index and measurement reliability coefficient, determine the retest priority of each detection area, and divide each detection area into at least two categories: direct acceptance area, local encrypted rescanning area, and contact verification area.

[0068] The retest priority is used to rank the urgency of additional testing for each detection area, taking into account both regional importance and measurement reliability. The comprehensive retest priority score is a quantitative score obtained by multiplying the importance index by the inverse of the measurement reliability coefficient. A higher score indicates that the area is both important and unreliable, requiring priority retesting. The direct acceptance area refers to areas with high measurement reliability coefficients and low retest priority scores; the initial detection results for these areas can be directly used as the final output. The locally encrypted rescan area represents areas with moderate measurement reliability coefficients and local noise; these areas require non-contact encrypted scanning to improve point cloud density. The contact verification area refers to areas with low measurement reliability coefficients or extremely high functional importance; these areas require high-precision retesting using contact probes to obtain baseline measurement data.

[0069] After determining the measurement reliability coefficients of each testing area, the system needs to decide which areas require additional testing and what retesting method to use. Specifically, the importance index of each testing area is multiplied by the reciprocal of its measurement reliability coefficient to obtain a comprehensive retesting priority score. This calculation logic follows the principle that areas with high importance and low reliability should be retested first. The system presets a first priority threshold of 0.3 and a second priority threshold of 0.7, and the thresholds can be dynamically adjusted based on the testing cycle time budget and historical batch data.

[0070] If the overall score of the detection area is less than 0.3, it indicates that the area is either unimportant or the measurement is reliable. It is classified as a direct acceptance area, and the system will not perform additional detection on the area, but will directly use the corrected initial detection result.

[0071] If the overall score is between 0.3 and 0.7, it indicates that there are certain reliability issues in the area but they have not yet reached a serious level. This area is classified as a localized encrypted rescanning area. The system performs laser scanning with half the step size in this area to increase the sampling point density and reduce noise.

[0072] If the overall score is greater than or equal to 0.7, it indicates that the reliability of the area is seriously insufficient or the function is extremely critical. This area is designated as a contact verification area, and the system calls a contact coordinate measuring machine to perform high-precision detection on the key feature points of this area. When the detection cycle is limited, the system can sort the areas by score from high to low, prioritizing the retesting of the top 30% of the areas, while the remaining areas are downgraded to encrypted rescanning or direct acceptance.

[0073] In some preferred embodiments, the above steps specifically include: multiplying the importance index of each detection area by the reciprocal of the measurement reliability coefficient to obtain the comprehensive retest priority score of each detection area as the retest priority; setting a first priority threshold and a second priority threshold, wherein the first priority threshold is less than the second priority threshold; if the comprehensive retest priority score of a detection area is less than the first priority threshold, then it is classified as a direct acceptance area.

[0074] If the overall retest priority score of the detection area is greater than or equal to the first priority threshold and less than the second priority threshold, it is classified as a local encrypted rescanning area; if the overall retest priority score of the detection area is greater than or equal to the second priority threshold, it is classified as a contact verification area.

[0075] The comprehensive retest priority score is a quantitative score obtained by multiplying the importance index by the reciprocal of the measurement reliability coefficient. The calculation formula is: Comprehensive retest priority score equals importance index divided by measurement reliability coefficient. The first and second priority thresholds represent the dividing values ​​for the three types of detection areas. These thresholds can be set using a fixed value method or an adaptive dynamic adjustment method. In the fixed value setting method, the first priority threshold is set to 0.3, and the second priority threshold is set to 0.7. In the adaptive dynamic adjustment method, the threshold is determined based on the quantile of the comprehensive score of all detection areas in the current batch. After sorting the scores from low to high, the upper limit of the scores of the top 40% of areas is used as the first priority threshold, and the upper limit of the scores of the top 70% of areas is used as the second priority threshold. Detection actions triggered by the direct acceptance area do not perform additional detection and directly output the corrected initial detection result. Detection actions triggered by the local encrypted rescan area perform a non-contact laser encrypted scan with half the scanning step distance in that area. Detection actions triggered by the contact verification area call a contact coordinate measuring machine to perform high-precision detection of key feature points in that area.

[0076] After determining the measurement reliability coefficient of each testing area, the system needs to decide on a retesting resource allocation strategy. Specifically, the importance index of each testing area is multiplied by the reciprocal of its measurement reliability coefficient to obtain a comprehensive retesting priority score. This calculation logic reflects the principle that high-importance, low-reliability areas should be retested first. The system adopts an adaptive threshold setting method, sorting the comprehensive scores of the 56 testing areas in the current batch from low to high. The score of the 22nd area (0.31) is used as the first priority threshold, and the score of the 39th area (0.68) is used as the second priority threshold.

[0077] If the overall score of the detected area is less than 0.31, it is determined to be a directly accepted area, and the system will not perform additional detection on this type of area. If the overall score is between 0.31 and 0.68, it is determined to be a locally encrypted rescanning area, and the system performs encrypted scanning on this type of area with the scanning step size reduced from 0.2 mm to 0.1 mm. If the overall score is greater than or equal to 0.68, it is determined to be a contact verification area, and the system calls a contact probe to perform contact measurements on the tooth surface meshing line, the center of the reference hole, and the key feature points of the end face flatness in this type of area. When the detection cycle is limited to completing all detections within 1 hour, the system sorts the areas by overall score from high to low, prioritizing the top 15 areas to complete contact verification, and downgrading the remaining high-scoring areas to locally encrypted rescanning.

[0078] S107. Perform additional detection on the local encrypted rescanning area and / or contact verification area to obtain additional detection results.

[0079] The supplementary detection refers to the second round of measurement operations performed by the system on the local encrypted rescanning area and the contact verification area. This detection selects either non-contact encrypted scanning or contact detection method according to the area type. The supplementary detection results are used to represent the additional measurement data obtained by the supplementary detection, including encrypted point cloud datasets or contact measurement point coordinate sets.

[0080] After completing the classification of the detection areas, the system initiates an additional detection process to perform refined measurements on the areas requiring re-testing. Specifically, for the locally intensified re-scanning area, the system re-plans the scanning path for that area, sets the scanning step size to 50% of the initial detection, increases the laser power by 10% to 15% to enhance the signal-to-noise ratio, reduces the scanning speed to 70% of the initial speed to improve single-point sampling time, performs a second scan on the area, and acquires point cloud data with a density increased by 4 times.

[0081] For the contact verification area, the system invokes a contact probe to extract the locations of key feature points based on the CAD model of the area, including sampling points for the tooth meshing line, the center of the reference hole, and the end face flatness measurement points. The probe approaches the workpiece surface at a speed of 0.1 mm / s and triggers the measurement. Each feature point is measured three times repeatedly, and the average value is taken to eliminate random errors. The retest range is automatically extended 5 to 10 mm beyond the area boundary to ensure the continuity of boundary data. If the standard deviation of the point cloud noise in the area still exceeds 0.05 mm or the contact measurement repeatability exceeds 0.01 mm after retesting, the system triggers an anomaly reporting process, suspends automatic detection, and notifies manual verification. The operator can choose to adjust the clamping state and re-detect or mark the area as a defect area and record it to the quality traceability system.

[0082] S108. Based on the deviation information, perform regional-level correction on the initial detection results and the additional detection results to obtain the corrected initial detection results and the corrected additional detection results.

[0083] The global spatial error compensation matrix is ​​the homogeneous transformation matrix calculated based on the deviation between the current pose and the nominal pose of the reference component. The region-level attenuation factor represents the nonlinear mapping weight of the spatial distance from the centroid of the detection region to the reference datum. The translation error component and rotation error component refer to the spatial translation vector and rotation submatrix decomposed from the global matrix. The local transformation matrix represents the reconstructed region-level correction matrix after nonlinear scaling by the attenuation factor.

[0084] It should be noted that in the inspection of large precision parts, the deviation information to be compensated is not, in essence, the rigid body pose error of the part itself, but rather a spatially non-uniform measurement error field formed by the superposition of multiple sources of disturbances, including the non-uniform thermal expansion field caused by temperature gradient, the non-stationary pose drift caused by micro-vibration of the inspection platform, the local strain caused by elastic deformation of the clamping system, sensor zero drift, and the deflection due to the weight of the large part. This error field has a non-uniform distribution characteristic in space, with the location of the reference part or reference feature as the observable origin, and decays with spatial distance and disturbance mode. Therefore, it cannot be compensated by applying a global uniform transformation to the rigid body.

[0085] If the global spatial error compensation matrix obtained from the reference part deviation is directly applied as a unified rigid body transformation to the entire detection area, two types of problems will occur: First, undercompensation will occur in areas close to the reference point because the actual deviation in this area is smaller than the deviation at the reference point but is compensated by an equal amount; second, severe overcompensation will occur in distant areas far from the reference point, especially since the rotational error component is linearly amplified into a positional error far exceeding the actual physical deviation under a long lever arm, which degrades the accuracy of the original measurement data. Therefore, the "local transformation matrix" described in this embodiment is not used to correct the rigid body pose of the part, but rather to invert the true amount of contamination of the measurement results by the local error field in each detection area and provide inverse compensation.

[0086] This step is triggered when acquiring deviation information and error compensation is needed for the original point cloud. Specifically, the system constructs a four-row, four-column global spatial error compensation matrix based on the reference component deviation. For each detection region, the spatial distance from the centroid of that region to the reference component is calculated, and a region-level attenuation factor is generated using an exponential decay function. The global spatial error compensation matrix is ​​decomposed into translational error components and rotational error components. The translational and rotational error components are nonlinearly scaled using this region-level attenuation factor. This nonlinear scaling aims to suppress the leverage effect of rotational error amplification with spatial distance, avoiding large positional shifts in distant regions due to small angular deviations. The local transformation matrix is ​​reconstructed based on the scaled components, and rigid body transformation is performed on the 3D coordinates in the initial and additional detection results using this local transformation matrix. In multi-reference collaborative scenarios, multi-reference weighted fusion correction can be introduced to reserve a data interface for subsequent thermal deformation field compensation.

[0087] In some preferred embodiments, the above steps specifically include: constructing a global spatial error compensation matrix based on the deviation information; for each detection area, calculating the spatial distance from the centroid of the detection area to the reference part or reference feature, and generating a region-level attenuation factor based on the spatial distance; and decomposing the global spatial error compensation matrix into translation error components and rotation error components.

[0088] Differential nonlinear scaling of the translation error component and the rotation error component is performed using a region-level attenuation factor. Wherein, the translation error component t = (t... x ,t y ,t z The three components of ) are each multiplied by a first-order power decay factor α(d) = e −d / L0 (Where d is the Euclidean distance from the centroid of the detection area to the reference piece, and L0 is the attenuation constant; in this embodiment, L0 = 200 mm), the local translation error component t is obtained. local =α(d)·t; The rotation error component is in the equivalent rotation vector θ=(θ x ,θy ,θ z After characterization, the rotation angle is multiplied by the square of the attenuation factor to obtain the local rotation error component θ. local =α 2 (d)·θ.

[0089] The physical basis for using different power-law attenuation coefficients for translation and rotation components is as follows: In three-dimensional space, the contribution of rotation error to the actual position of a point at a distance dd from the rotation center is approximately d⋅sin(Δθ)≈d·Δθ, meaning that the rotation error itself implicitly includes a lever amplification of spatial distance. If only a power-law attenuation is applied to the rotation angle, the compensated equivalent position error still increases linearly with distance, failing to eliminate the lever effect. However, with a quadratic power-law attenuation, the equivalent contribution of rotation error to position is d⋅α²(d)⋅Δθ. Since α(d) itself decreases monotonically with d, the overall expression is effectively suppressed in the far-end region, thus inhibiting the lever amplification effect of rotation error. Those skilled in the art can adjust the attenuation power of the rotation component between 1.5 and 2.5 according to the specific part dimensions, sensor arrangement, and historical calibration data to adapt to different working conditions.

[0090] The translation error component does not suffer from the leverage amplification problem mentioned above. Its physical influence remains essentially constant with spatial distance (only slowly decaying due to non-uniformity of the thermal or elastic field). Therefore, a first-order power decay is sufficient to meet the compensation requirements. This differentiated power-order design ensures that the spatial distribution of the compensation intensity matches the physical propagation law of the actual error field.

[0091] Based on the local translation error components and local rotation error components, the local transformation matrix for each detection region is reconstructed; using the local transformation matrix, rigid body transformations are performed on the initial detection results and the additional detection results respectively to obtain the corrected initial detection results and the corrected additional detection results.

[0092] The global spatial error compensation matrix is ​​a four-row, four-column homogeneous transformation matrix calculated using least squares or singular value decomposition based on the deviation between the current pose and the nominal pose of the reference or reference feature. This matrix describes the spatial error of the overall measurement coordinate system relative to the theoretical coordinate system. The region-level attenuation factor represents the spatial correlation weight of each detection region relative to the reference datum. This factor is a value between 0 and 1, decreasing as the spatial distance from the centroid of the detection region to the reference or reference feature increases. The translation error component represents the three-dimensional vector describing the translation of the origin in the global spatial error compensation matrix, in millimeters. The rotation error component represents the three-row, three-column rotation submatrix or equivalent rotation vector describing the rotation of the coordinate axes in the global spatial error compensation matrix, in radians. The local transformation matrix represents the four-row, four-column homogeneous transformation matrix reconstructed for a specific detection region. This matrix integrates the local translation error component and the local rotation error component after scaling by the region-level attenuation factor.

[0093] After acquiring the deviation information of the reference part or reference feature, the system maps the deviation to a local correction matrix for each detection area. Specifically, based on the current measured pose and nominal pose of the reference part, the optimal rigid body transformation is solved using the singular value decomposition method, constructing a four-row, four-column global spatial error compensation matrix. For each detection area, the Euclidean distance between the centroid coordinates of the point cloud of that detection area and the center of the reference part or reference feature is calculated. This distance is then substituted into an exponential decay function to generate a regional attenuation factor. The attenuation function is in the form that the regional attenuation factor equals the exponentially negative distance divided by 200 mm. When the distance is 0, the attenuation factor is 1.0; when the distance is 200 mm, the attenuation factor is 0.37. The global spatial error compensation matrix is ​​decomposed into translation error components and rotation error components. The translation error components are extracted from the first three rows of the fourth column of the matrix, and the rotation error components are extracted from the three rows and three columns of the upper left submatrix of the matrix and converted into rotation vectors.

[0094] By multiplying the three components of the translation error component by a regional attenuation factor and the rotation angle of the rotation error component by the square of the attenuation factor, this nonlinear scaling process suppresses the leverage effect of rotation error amplification with spatial distance, resulting in local translation and rotation error components. Based on these components, a four-row, four-column local transformation matrix for the detection region is reconstructed. Using this matrix, homogeneous coordinate transformations are performed on all point cloud coordinates in the region in both the initial and subsequent detection results, completing rigid body correction and yielding corrected initial and subsequent detection results.

[0095] S109. Perform replacement processing or weighted fusion processing on the corrected initial test results and the corrected additional test results to output the final test results of the large precision parts to be tested.

[0096] Point cloud registration refers to the spatial transformation process of aligning heterogeneous measurement data to a unified coordinate system. Sampling point density distribution represents the number of effective measurement points per unit area. Forced constraint deformation refers to the operation of driving local point cloud deformation and fitting using high-precision contact measurement points as hard constraint boundaries. Confidence weight represents the mesh node fusion coefficient constructed based on the measurement reliability coefficient. Weighted distance inverse interpolation algorithm is a numerical analysis method that calculates the coordinates of transition zone nodes based on the inverse of spatial distance and confidence weight.

[0097] This step is executed after data correction for each region is completed and a full-size 3D topographic model needs to be generated. Specifically, for the directly accepted area, the system directly outputs the corrected initial detection results. For the locally encrypted rescanned area, the corrected additional detection results are iteratively registered with the initial detection results, and then weighted and averaged according to the sampling point density distribution ratio. For the contact verification area, the contact measurement points of the additional detection are extracted as hard constraint benchmarks, and the local point cloud of the initial detection results is subjected to forced constraint deformation using Laplacian mesh deformation, followed by curvature adaptive resampling to complete data replacement. At the boundary edges of each region, the measurement reliability coefficients of sampling points in adjacent regions are extracted to construct confidence weights, and a weighted inverse distance interpolation algorithm is used to smoothly stitch the transition areas into a grid. This processing logic stitches discrete local optimal solutions into a globally continuous high-fidelity model, and finally synchronously generates the final detection results containing size deviations and regional confidence maps, providing a complete chain of evidence for quality traceability.

[0098] In some preferred embodiments, the above steps specifically include: for the direct acquisition area, directly outputting the corrected initial detection result as the final detection result; for the local encrypted rescanning area, performing point cloud registration between the corrected additional detection result and the corrected initial detection result, and performing weighted average fusion based on the sampling point density distribution of the two to obtain the final detection result;

[0099] For the contact verification area, the contact measurement points of the corrected additional detection results are used as the benchmark. The local point cloud of the corresponding corrected initial detection results is subjected to forced constraint deformation and resampling to complete the data replacement. At the boundary edge of the direct acceptance area, the local encrypted rescanning area and the contact verification area, the confidence weight is constructed based on the measurement reliability coefficient of the sampling points in the adjacent areas. The data in the transition area is meshed and stitched using a weighted distance inverse interpolation algorithm to output the final detection result.

[0100] After obtaining the corrected initial detection results and the supplementary detection results, the system executes a differentiated data fusion strategy based on the region type. Specifically, for directly accepted regions, the corrected initial detection results are directly output as the final detection results, without performing any fusion operations.

[0101] For the locally encrypted rescanned area, the corrected additional detection point cloud and the corrected initial detection point cloud are registered to a unified coordinate system using an iterative nearest point algorithm. The sampling point density distribution of the two sets of point clouds is calculated. The initial detection point cloud has a density of 4 points per square millimeter, and the additional detection point cloud has a density of 16 points per square millimeter. Based on the density ratio, a weight of 0.2 is assigned to the initial detection point cloud and a weight of 0.8 is assigned to the additional detection point cloud. The corresponding point coordinates of the two sets of point clouds are then weighted and averaged to obtain the final detection result point cloud.

[0102] For the contact verification area, the coordinates of the contact measurement points in the corrected additional detection results are extracted. These coordinates are used as hard constraint points. Laplacian mesh deformation is performed on the local point cloud of this area in the corrected initial detection results to make the point cloud strictly match the position of the contact measurement points. After the forced constraint deformation is completed, the deformed point cloud is resampled according to the curvature adaptive resampling at a density of 8 points per square millimeter to complete the data replacement.

[0103] In the transition region with a width of 10 to 15 mm at the boundary between the direct acceptance area, the local encrypted rescanning area, and the contact verification area, the measurement reliability coefficient of the sampling points at the boundary of the adjacent area is extracted as the confidence weight. For the interpolation point in the transition region, the Euclidean distance from the point to each sampling point in the adjacent area is calculated. The weighted inverse distance interpolation algorithm is adopted, and the interpolation weight is set as the normalized value of the product of the inverse distance and the confidence weight. The coordinates of the interpolation point are calculated, and the grid stitching of the transition region is realized. Finally, the complete final detection result containing all regions is output.

[0104] The following describes the adaptive detection equipment for machining accuracy of large precision parts in the embodiments of this invention from the perspective of hardware processing. Please refer to [link to relevant documentation]. Figure 2 This is a schematic diagram of the physical device structure of an adaptive detection device for machining accuracy of large precision parts in this application embodiment.

[0105] It should be noted that, Figure 2 The structure of the large precision parts machining accuracy adaptive inspection equipment shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of the present invention.

[0106] like Figure 2As shown, the adaptive inspection device for machining accuracy of large precision parts includes a central processing unit (CPU) 201, which can perform various appropriate actions and processes based on programs stored in read-only memory (ROM) 202 or programs loaded from storage section 208 into random access memory (RAM) 203, such as executing the methods described in the above embodiments. The RAM 203 also stores various programs and data required for system operation. The CPU 201, ROM 202, and RAM 203 are interconnected via bus 204. An input / output (I / O) interface 205 is also connected to bus 204.

[0107] The following components are connected to I / O interface 205: input section 206 including audio input devices, push-button switches, etc.; output section 207 including a liquid crystal display (LCD) and audio output devices, indicator lights, etc.; storage section 208 including a hard disk, etc.; and communication section 209 including a network interface card such as a LAN (Local Area Network) card, modem, etc. Communication section 209 performs communication processing via a network such as the Internet. Drive 210 is also connected to I / O interface 205 as needed. Removable media 211, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., are installed on drive 210 as needed so that computer programs read from them can be installed into storage section 208 as needed.

[0108] In particular, according to embodiments of the present invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of the present invention include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing computer programs for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication section 209, and / or installed from removable medium 211. When the computer program is executed by central processing unit (CPU) 201, it performs the various functions defined in the present invention.

[0109] It should be noted that specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), flash memory, optical fiber, portable compact disc read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this invention, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0110] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. Each block in a flowchart or block diagram may represent a module, program segment, or portion of code, which contains one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those shown in the drawings.

[0111] Specifically, the adaptive detection device for machining accuracy of large precision parts in this embodiment includes a processor and a memory. The memory stores a computer program, and when the computer program is executed by the processor, it implements the adaptive detection method for machining accuracy of large precision parts provided in the above embodiment.

[0112] In another aspect, the present invention also provides a computer-readable storage medium, which may be included in the adaptive detection device for machining accuracy of large precision parts described in the above embodiments; or it may exist independently and not assembled into the adaptive detection device for machining accuracy of large precision parts. The storage medium carries one or more computer programs, which, when executed by a processor of the adaptive detection device for machining accuracy of large precision parts, cause the adaptive detection device for machining accuracy of large precision parts to implement the adaptive detection method for machining accuracy of large precision parts provided in the above embodiments.

[0113] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit it. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application.

[0114] As used in the above embodiments, depending on the context, the term "when..." can be interpreted as meaning "if...", "after...", "in response to determining...", or "in response to detecting...". Similarly, depending on the context, the phrase "when determining..." or "if (the stated condition or event) is interpreted as meaning "if determining...", "in response to determining...", "when (the stated condition or event) is detected", or "in response to detecting (the stated condition or event)".

[0115] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. This program can be stored in a computer-readable storage medium, and when executed, it can include the processes described in the above method embodiments. The aforementioned storage medium includes various media capable of storing program code, such as ROM or random access memory (RAM), magnetic disks, or optical disks.

Claims

1. An adaptive detection method for machining accuracy of large precision parts, characterized in that, The method includes: Obtain the model information and testing requirements of the large precision part to be tested, and divide the large precision part to be tested into multiple testing areas; The importance indicators of each detection area are determined based on the model information and the detection requirements; A reference component or reference feature is set on the unmachined reference surface of the large precision part to be tested or on its clamping fixture; An initial inspection is performed on the large precision part to be tested to obtain an initial inspection result, and state data is collected. The state data includes deviation information of the reference part or the reference feature relative to the nominal state. Based on the state data, environmental disturbance and clamping stability features are extracted, and combined with the data quality features of the initial detection results, multi-dimensional evaluation sub-indicators are constructed. After normalizing the evaluation sub-indicators, the measurement reliability coefficient of each detection area is determined. Based on the importance index and the measurement reliability coefficient, the retest priority of each detection area is determined, and each detection area is divided into at least two categories: direct acceptance area, local encrypted rescan area, and contact verification area. Additional detection is performed on the local encrypted rescanning area and / or the contact verification area to obtain additional detection results; Based on the deviation information, the initial detection result and the additional detection result are corrected at the region level to obtain the corrected initial detection result and the corrected additional detection result; The modified initial detection results and the modified additional detection results are subjected to replacement processing or weighted fusion processing to output the final detection results of the large precision part to be tested.

2. The method according to claim 1, characterized in that, The process involves extracting environmental disturbance and clamping stability features based on the state data, and combining this with the data quality features of the initial detection results to construct multi-dimensional evaluation sub-indicators. After normalizing these evaluation sub-indicators, the measurement reliability coefficients for each detection area are determined. Specifically, this includes: Analyze the state data, extract the temperature gradient change rate and platform vibration amplitude, and calculate the environmental disturbance degree sub-indices; Extract the displacement of the clamping point of the large precision part to be tested before and after the test, and calculate the clamping stability sub-index. The deviation information is compared with the historical benchmark deviation to calculate the reference consistency sub-index; Based on the point cloud noise distribution and local curvature variance of each detection region in the initial detection results, the initial detection consistency sub-index is calculated. Preset weights are assigned to the sub-indicators, and the weighted summation results are normalized to obtain the measurement reliability coefficients of each detection area. The measurement reliability coefficients are negatively correlated with the temperature gradient change rate, the platform vibration amplitude, the displacement, and the point cloud noise distribution.

3. The method according to claim 2, characterized in that, The step of extracting the displacement of the clamping point of the large precision part under test before and after testing, and calculating the clamping stability sub-index, specifically includes: The coordinate sequence of the clamping point is collected in real time during the initial detection process by using an optical target or displacement sensor arranged at the clamping fixture. The maximum offset of the coordinate sequence is calculated as the displacement. If the maximum offset exceeds a preset elastic deformation threshold, it is determined that the clamping has become loose and the target clamping point where the clamping has become loose is determined. For the target clamping point where clamping loosening occurs, a spatial penalty function with distance decay centered on the target clamping point is constructed; The spatial penalty function is used to penalize and amplify the clamping stability sub-index of the detection area adjacent to the target clamping point, so as to reduce the measurement reliability coefficient of the corresponding detection area.

4. The method according to claim 1, characterized in that, Based on the importance index and the measurement reliability coefficient, the retesting priority of each detection area is determined, and each detection area is divided into at least two categories: direct acceptance area, local encrypted rescanning area, and contact verification area, specifically including: Multiply the importance index of each detection area by the reciprocal of the measurement reliability coefficient to obtain the comprehensive retest priority score of each detection area, which is used as the retest priority. Set a first priority threshold and a second priority threshold, wherein the first priority threshold is less than the second priority threshold; If the overall retest priority score of the detection area is less than the first priority threshold, then it is classified as the direct acceptance area; If the overall retest priority score of the detection area is greater than or equal to the first priority threshold and less than the second priority threshold, then it is classified as the local encrypted rescan area. If the overall retest priority score of the detection area is greater than or equal to the second priority threshold, then it is classified as the contact verification area.

5. The method according to claim 1, characterized in that, The step of performing region-level correction on the initial detection result and the additional detection result based on the deviation information to obtain the corrected initial detection result and the corrected additional detection result specifically includes: Based on the deviation information, a global spatial error compensation matrix is ​​constructed; For each of the detection regions, the spatial distance from the centroid of the detection region to the reference element or the reference feature is calculated, and a region-level attenuation factor is generated based on the spatial distance. The global spatial error compensation matrix is ​​decomposed into translation error components and rotation error components; The translation error component and the rotation error component are nonlinearly scaled using the regional attenuation factor to suppress the leverage effect of rotation error amplification with spatial distance, thereby obtaining local translation error component and local rotation error component. Based on the local translation error components and the local rotation error components, the local transformation matrix for each detection region is reconstructed; Using the local transformation matrix, rigid body transformations are performed on the initial detection result and the additional detection result to obtain the corrected initial detection result and the corrected additional detection result.

6. The method according to claim 1, characterized in that, The process of performing replacement processing or weighted fusion processing on the corrected initial detection results and the corrected additional detection results to output the final detection results of the large precision part to be tested specifically includes: For the direct acceptance region, the corrected initial detection result is directly output as the final detection result; For the local encrypted rescanning area, the corrected additional detection result and the corrected initial detection result are registered in point cloud, and a weighted average fusion is performed based on the sampling point density distribution of the two to obtain the final detection result; For the contact verification area, based on the contact measurement points of the corrected additional detection results, the local point cloud of the corresponding corrected initial detection results is subjected to forced constraint deformation and resampling to complete the data replacement; At the boundary edge of the direct acceptance area, the local encrypted rescanning area, and the contact verification area, confidence weights are constructed based on the measurement reliability coefficients of sampling points in adjacent areas. A weighted distance inverse interpolation algorithm is used to perform gridded stitching of the data in the transition area, and the final detection result is output.

7. The method according to claim 1, characterized in that, Before the steps of acquiring the model information and inspection requirements of the large precision part to be tested, and dividing the large precision part to be tested into multiple inspection areas, the method further includes: The model information of the large precision part to be tested is used to perform topological structure and mass distribution analysis; Multiple rigid nodes in the large precision part to be tested were identified; The location of the rigid node is set as the mounting position of the reference component, or the inherent geometric features at the rigid node are extracted as the reference features.

8. An adaptive detection device for the machining accuracy of large precision parts, characterized in that, The adaptive detection device for machining accuracy of large precision parts includes: one or more processors and a memory; the memory is coupled to the one or more processors, the memory is used to store computer program code, the computer program code includes computer instructions, and the one or more processors call the computer instructions to cause the adaptive detection device for machining accuracy of large precision parts to perform the method as described in any one of claims 1-7.

9. A computer-readable storage medium comprising instructions, characterized in that, When the instruction is executed on a large precision parts machining accuracy adaptive inspection device, the large precision parts machining accuracy adaptive inspection device performs the method as described in any one of claims 1-7.

10. A computer program product, characterized in that, When the computer program product is run on a large precision parts machining accuracy adaptive detection device, the large precision parts machining accuracy adaptive detection device performs the method as described in any one of claims 1-7.

Images