Bearing production quality detection system based on production data analysis

Abstract

The application discloses a bearing production quality detection system based on production data analysis, which comprises five modules of data acquisition, multi-process residual error coupling and decoupling, adaptive threshold quality inspection, data storage and result output. Residual errors of forging, heat treatment, grinding and superfinishing processes and three types of individual process data of raw materials, tooling and machine tools are collected and standardized. The coupling and decoupling module dynamically calibrates process coupling coefficients based on covariance and discrete entropy, calculates independent and cross-coupling hazard degrees in layers, and quantifies the risk of hidden defects formed by the superposition of compliance errors. The quality inspection module constructs a process endowment index, dynamically generates individualized judgment thresholds in linkage with the coupling hazard degree, divides four levels of evaluation of high quality / qualified / early warning / unqualified, and judges data to iteratively optimize algorithm parameters. The whole link data is archived and visualized to output quality inspection and process optimization reports. The application overcomes the problems of traditional single-process and fixed threshold detection, such as missing hidden defects and misjudging product defects.

Description

Technical Field

[0001] This invention relates to the field of bearing production quality inspection, specifically a bearing production quality inspection system based on production data analysis. Background Technology

[0002] Bearings are core components in high-end equipment manufacturing. Their production involves a multi-stage composite forming process, mainly including four core processes: forging, heat treatment and tempering, precision grinding, and ultra-precision polishing. When the minute residual errors from any single process are within the national standard's acceptable threshold, traditional quality inspection techniques would directly classify the bearing as a qualified product. However, in large-scale industrial production, existing bearing quality inspection technologies have the following technical problems, specifically: First, the coupling and superposition of compliant residual errors from multiple processes can lead to latent service defects, posing a serious risk of missed detection. Each manufacturing process of bearings inevitably produces minute residual errors: forging introduces minor residual deviations in material density, heat treatment introduces minor residual stress deformation, grinding introduces micron-level dimensional residual deviations, and ultra-precision machining introduces microscopic roughness residual errors. Existing quality inspection technologies all employ independent threshold detection mechanisms for each process, only judging whether the residual error of each process meets the standards, ignoring the coupling and superposition effect of residual errors from multiple processes. In actual production, bearings where the residual errors of individual processes are all within the acceptable range can experience multi-dimensional errors that couple and superimpose in the raceway and ball contact area, resulting in latent defects such as localized stress concentration and contact accuracy deviations. These bearings may pass factory quality inspection, but under high-speed, heavy-load service conditions, the coupled residual errors will continue to amplify, ultimately leading to abnormal noise, wear, jamming, or even fracture failure. Currently, the industry lacks decoupling analysis and latent defect identification technologies for compliant residual errors from multiple processes, making it impossible to quantify the degree of error coupling, which is a core data challenge for early failure of high-end bearings.

[0003] Secondly, individual differences in the manufacturing processes of bearings within the same batch lead to distortions in the determination of uniform static quality inspection thresholds. Bearings produced on the same production line and in the same batch are affected by minute differences in the crystal structure of raw materials, fluctuations in machine tool instantaneous processing stiffness, and slight deviations in tooling clamping. Each bearing possesses its own unique individual manufacturing process characteristics, meaning that its error tolerance threshold and structural stability exhibit natural individual differences. Existing quality inspection technologies all use uniform, fixed thresholds for quality judgment, failing to differentiate between these individual characteristics. For bearings with excellent manufacturing processes and high error tolerance, slightly exceeding process tolerances will not affect service performance, but a uniform threshold will misjudge them as defective, resulting in wasted production resources. For bearings with weak manufacturing processes and poor structural stability, although the errors in each process may be within a uniform acceptable threshold, the accumulation of small errors can lead to service failure, and a uniform threshold will misjudge them as acceptable, causing defective products to enter the market. This problem is a unique pain point in the mass production of bearings across multiple processes, and general quality inspection algorithms cannot adapt to the needs of individualized testing.

[0004] In summary, existing bearing quality inspection technologies are detached from the multi-process composite molding production nature and lack effective solutions for bearing-specific error coupling and individual endowment differences. They are unable to meet the high-precision and high-reliability quality inspection requirements of high-end bearings, and there is an urgent need to design a bearing production quality inspection system based on production data analysis. Summary of the Invention

[0005] The purpose of this invention is to provide a bearing production quality inspection system based on production data analysis to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a bearing production quality inspection system based on production data analysis, comprising a data acquisition module, a multi-process residual error coupling and decoupling feature extraction module, an individual process endowment adaptive threshold quality inspection judgment module, a data storage module, and a result output module; the data acquisition module collects two types of data sources: the first type is multi-process forming residual error data, including four types of original error parameters: forging process density residual deviation, heat treatment process stress residual deformation, grinding process dimensional residual deviation, and ultra-precision process roughness residual deviation; the second type is individual process endowment microscopic data, including three types of original process parameter data: raw material crystal phase uniformity, tooling clamping pressure fluctuation value, and machine tool instantaneous machining accuracy; the collected data is then cleaned, normalized, and standardized to obtain preprocessed data; The multi-process residual error coupling and decoupling feature extraction module is used to receive the preprocessed data from the acquisition module and collect the residual error data of bearing forging, heat treatment, grinding and ultra-precision multi-process. Through the multi-dimensional residual error hierarchical coupling and decoupling algorithm, the module realizes the hierarchical decoupling of independent errors of single process and coupled errors of multiple processes, quantifies the total coupling hazard of errors, and extracts the features of hidden defects. The individual process endowment adaptive threshold quality inspection judgment module is linked with the multi-process residual error coupling and decoupling feature extraction module. Based on the quantified error coupling hazard degree and combined with the individual micro-process parameters of the bearing in the preprocessed data, the module realizes personalized quality inspection threshold correction and differentiated quality judgment for a single bearing through the process endowment difference dynamic threshold correction algorithm. The data storage module is used to receive and store the full-link data from the data acquisition module, the multi-process residual error coupling and decoupling feature extraction module, the individual process endowment adaptive threshold quality inspection judgment module, and the result output module. The result output module is used to receive the final data from the data storage module and the individual process endowment adaptive threshold quality inspection judgment module. Based on the calculation results of the coupling decoupling algorithm and the dynamic threshold correction algorithm, it completes the standardized integration, hierarchical output, working condition traceability display and batch optimization result push of bearing quality inspection data.

[0007] Preferably, the specific implementation steps of the multi-process residual error coupling decoupling feature extraction module are as follows: Step A1: Collection and Standardization of Residual Error Data from Multiple Processes: Receive preprocessed data from the acquisition module and collect four categories of original residual error data: forging density deviation of a single bearing, residual deformation from heat treatment, grinding dimensional deviation, and ultra-precision roughness deviation. Remove abnormal extreme values ​​caused by equipment failures. Use an extreme value normalization method to uniformly transform process error data of different dimensions and ranges into standardized residual errors within the 0~1 range. Eliminate the impact of dimensional differences on coupling analysis; Step A2, Statistical Analysis of Process Error Correlation and Coupling Coefficient Calculation: Based on the multi-process error dataset of bearings in the same batch, calculate the covariance and variance of each pair of process errors, substitute them into the coupling coefficient calculation formula, and obtain the error coupling influence coefficient between any two processes. Combined with bearing model matching coupling correction coefficient This allows for adaptive calibration of the coupling coefficient, precisely matching the forming characteristics of different bearings. Step A3, Layered Decoupling and Coupling Hazard Calculation: Using the coupling decoupling formula, the independent error impact value of a single process and the cross-coupling error impact value of two processes are calculated layer by layer, and finally the total coupling hazard of a single bearing is obtained by summing them up. ; It can accurately quantify the hidden defect risks caused by the superposition of compliance errors in multiple processes, making up for the shortcomings of traditional single-process inspection in that it cannot identify coupled defects; Step A4, Feature Output and Data Transmission: The calculated total coupling hazard degree... The independent error and cross-coupling error characteristic data of each process are packaged and output to the individual process endowment adaptive threshold quality inspection judgment module, providing accurate quantitative basis for individual process endowment threshold correction and differentiated quality judgment.

[0008] Preferably, the specific implementation steps for the statistical analysis of process error correlation and the calculation of coupling coefficient in step A2 are as follows: Step A21: Construction of multi-process error pairing dataset: Standardize residual errors based on the four core processes of a single bearing. forging, Heat treatment Grinding, Ultra-precision is the basic unit. It collects all sample data from the current complete production batch, constructs a pairwise paired error dataset for each process, and forms... , , , , , There are six sets of process pairings, covering all cross-process error coupling scenarios, providing complete data support for subsequent correlation calculations; Step A22, Solving for statistical characteristic quantities of process error: For each pair of processes... Calculate the covariance of process errors and the variance of single-process errors separately, specifically by setting up statistical rules: solve for the covariance of the two sets of process errors. , used to characterize the , The degree of correlation between residual errors in two processes is considered; a larger value indicates a stronger coupling and correlation between the errors of the two processes. The variance of single-process errors is calculated simultaneously. , It is used to characterize the degree of fluctuation and dispersion of residual error in a single process in batch production, and to reflect the stability of single-process processing; Step A23, Initial Coupling Correlation Coefficient Calculation: Substitute the covariance and variance data of each process group into the uncalibrated basic coupling coefficient formula to calculate the initial coupling correlation coefficient: ;in The initial coupling coefficient, which is not adapted to the bearing process characteristics, only reflects the correlation of process error from the perspective of data statistics, without distinguishing the differential coupling hazards of different bearing structures and processing characteristics; Step A24, Bearing Model Adaptive Coefficient Calibration: Based on the bearing types of the current production line, match the process coupling correction coefficient. Complete adaptive calibration; Step A25, Final Coupling Influence Coefficient Solution and Consolidation: The initial coupling coefficients are then... Category Adjustment Factor Substitute into the coupling coefficient formula The final error coupling influence coefficients corresponding to the six pairs of process operations were calculated one by one. , The actual coupling hazard weights after the superposition of errors from different processes were quantified, and adaptive calibration of all process coupling parameters was completed. The solved parameters were then used to calculate the total... The parameters are output to step A3.

[0009] Preferably, the process coupling correction coefficient in step A24 The specific calculation logic is as follows: Step A24.1, Calculation of Discrete Entropy of Single-Process Error: To quantify the degree of process fluctuation disorder in the current batch production of each process, a discrete entropy model of process error is constructed to characterize the disordered fluctuation characteristics of residual error in a single process. The formula for calculating the discrete entropy of single-process error is as follows: , specific For the first The discrete entropy of the error of each process, where the process number is... These correspond to forging, heat treatment, grinding, and ultra-precision processes, respectively. For the first Standardization residual error of each process The probability distribution density across all samples in the current batch; The number of intervals for process error statistics; The larger the value, the more drastic the process fluctuations, and the higher the risk of small residual errors in the process coupling and superimposing to form hidden defects. Step A24.2, Batch Global Process Correlation Calculation: Based on the inherent process sequence logic of the four bearing processes, a batch global process correlation coefficient is constructed to quantify the overall linkage coupling strength of the multi-process errors in the entire batch. The calculation formula is as follows: ;in This is the batch-wide process correlation coefficient, with a value range of [value range missing]. The larger the value, the stronger the correlation between multiple processing steps and the more significant the superposition and coupling effect of process errors. Step A24.3, Dynamic Coupling Correction Coefficient Fusion Solution: Combining the discrete entropy of single-process fluctuation characteristics with the correlation degree of global process linkage characteristics, an adaptive fusion algorithm is constructed to dynamically solve the coupling correction coefficient adapted to the current batch production conditions. Core calculation formula: ;in The mean of the discrete entropy of the errors in the four processes satisfies This is used to characterize the overall process fluctuation level of the current batch of bearings; a constant of 1.0 serves as the baseline correction base to ensure the stability of the parameter baseline; [This is to achieve...] The value is dynamically and adaptively adjusted according to the operating conditions, and eventually stabilizes at a certain value. interval; Step A24.4, Adaptive Logic for Operating Conditions: When producing precision miniature bearings, experiencing drastic batch process fluctuations, and having strong process interrelationships, and Increase synchronously. Automatically approaches 1.2, appropriately amplifies the weight of process coupling errors, and accurately captures minute hidden coupling defects; when producing heavy-duty industrial bearings, with high batch process stability and weak process linkage, and Decrease Automatically approaches 0.9, weakens invalid slight coupling interference, avoids misjudgment caused by over-detection, and achieves adaptive matching under all working conditions.

[0010] Preferably, the specific implementation steps of step A3 are as follows: Step A31, Core Calculation Parameter Collection and Validity Verification: First, collect all the prerequisite parameters required for this step, complete the parameter consistency verification, and ensure calculation accuracy: standardize residual errors in four processes. , , , (Values ​​range from 0 to 1), Independent error influence coefficient for each process The dynamic coupling influence coefficient of the six process pairings (Adaptively corrected by discrete entropy and global correlation) (Obtained); invalid samples with abnormal parameters or missing data are removed to ensure that all input parameters are valid data for adaptive solution of the current bearing batch operating conditions; Step A32, Calculation of Independent Hazardous Component of Single Process Error: Based on the hierarchical decoupling concept, the independent hazard effect of residual error in a single process is first isolated. This component represents the fundamental hazard of a small residual error in a single process to bearing quality, without cross-process superposition effect. The calculation formula is as follows: ;in, This represents the total hazard component of independent errors in a single process. The preset single-process independent error influence coefficient is used to characterize the basic hazard weight of errors in forging, heat treatment, grinding, and ultra-precision single processes; This step is for the standardized residual error of the corresponding process. It can independently quantify the basic negative impact of each qualified residual error, solving the problem that traditional technology can only determine the error is qualified but cannot quantify the basic degree of harm. Step A33, Calculation of the Hazardous Component of Cross-Process Coupling Errors: Calculate the additional hazard component caused by the superposition of errors across processes. The calculation formula is as follows: In the formula, This refers to the component of errors caused by cross-coupling of multiple processes; The dynamic process coupling influence coefficient, which is adaptively solved in step A2, integrates the entropy value of the current batch process fluctuation and the process correlation characteristics, and can accurately adapt to real-time production conditions. This is the coupled superposition of residual errors from two processes, characterizing the synergistic effect of minute errors in different processes. This component specifically quantifies the hidden defects such as stress concentration and contact misalignment caused by the coupling of errors in multiple processes; Step A34: Calculation of Total Coupling Hazard: Combining the basic hazard components of a single process with the superimposed hazard components of multiple processes, the total coupling hazard is obtained, fully characterizing the overall defect risk of the bearing. Substituting this into the coupling decoupling formula: ;in The total coupling hazard of residual errors from multiple processes in the bearing is represented, and its value is greater than or equal to 0. The numerical classification logic is as follows: The closer it is to 0, the more uniform the errors in each process of the bearing are, the weaker the coupling and superposition effect is, and the less risk there is of hidden service defects. The higher the value, the more significant the cumulative effect of compliance errors in multiple processes, and the higher the risk of latent bearing failure.

[0011] Step A35, Latent Defect Risk Quantification and Calibration: Based on the solution obtained... Complete defect risk stratification and labeling, which differs from the traditional coarse judgment logic of judging a product as good as that of a single process passing: single process error meets the standard but... Bearings exceeding the adaptive threshold are identified as risk samples with latent coupling defects, enabling the identification of latent defects that traditional testing methods cannot cover, thus making up for the shortcomings of existing quality inspection technologies in missing detection.

[0012] Preferably, the specific implementation steps of the individual process endowment adaptive threshold quality inspection judgment module are as follows: Step B1, Individual Process Parameter Collection and Endowment Index Calculation: Collect microscopic process data from the entire production process of a single bearing: raw material crystal phase uniformity test data, machining tool clamping pressure fluctuation data, and machine tool instantaneous machining accuracy data. After standardizing and scoring these three types of data, substitute them into the endowment index calculation formula to obtain the process endowment index specific to each single bearing. It accurately characterizes an individual's tolerance to error. Step B2, Adaptive Quality Inspection Threshold Dynamic Correction: Retrieve Industry Standard Fixed Thresholds The coupling hazard degree transmitted by the multi-process residual error coupling decoupling feature extraction module is combined with the feature extraction module. The endowment index calculated in step B1 The personalized quality inspection threshold for a single bearing is calculated using a dynamic threshold correction formula. ; Specific correction logic: For high-quality bearings with high endowment index and low coupling hazard, the threshold is appropriately relaxed to avoid over-inspection caused by slight errors; for weak bearings with low endowment index and high coupling hazard, the threshold is appropriately tightened to intercept potential defective products in advance. Step B3, Differentiated Quality Grade Determination: The actual process error and coupling error of a single bearing are compared with the personalized correction threshold, and four levels of judgment standards are preset: high-quality product (error is far below the correction threshold), qualified product (error meets the correction threshold requirements), warning product (error is close to the correction threshold and there is potential risk), and unqualified product (error exceeds the correction threshold); compared with the traditional unified threshold judgment, precise differentiated quality inspection is achieved.

[0013] Step B4, Data Reverse Iteration and Result Output: Output the quality inspection judgment result of a single bearing, and simultaneously store the individual bearing endowment parameters, threshold correction data, and coupling error matching data. Iterate backwards to optimize the process coupling correction coefficient of the multi-process residual error coupling decoupling feature extraction module. Improve coupling error in subsequent batches Preferably, the specific implementation steps for collecting individual process parameters and calculating the endowment index in step B1 are as follows: Step B11: Precise Collection and Anomaly Filtering of Multi-Dimensional Microscopic Process Parameters: Three types of core original process data that cause individual bearing endowment differences within the same batch are specifically collected: 1) Original test data on the uniformity of raw material crystal phase, characterizing the structural uniformity and stability of the bearing substrate; 2) Data on the fluctuation of clamping pressure of machining fixtures, characterizing the machining stability deviation during assembly and clamping; 3) Data on the instantaneous machining accuracy of the machine tool, characterizing the fluctuation of equipment operating accuracy during the machining of a single bearing. Simultaneously, the data cleaning logic of the multi-process residual error coupling and decoupling feature extraction module is used to eliminate extreme abnormal data caused by equipment failure and human error, retaining effective sample data that truly reflects the production characteristics of a single bearing, ensuring the accuracy of subsequent endowment calculations. Step B12, Multidimensional Parameter Normalization and Standardization Scoring: Due to the inconsistency in the dimensions and numerical ranges of the three types of original process parameters, direct weighted calculation is not possible. Therefore, the original data is normalized to convert them into a unified standard. The standardized scores for each interval yielded the following: raw material phase uniformity score. Tooling clamping stability score Machine tool machining accuracy rating The scoring rules are uniformly adapted to the judgment logic of this invention: the closer the score value is to 1, the better the process performance in that dimension and the stronger the bearing error tolerance; the closer the score value is to 0, the more minor defects exist in the process in that dimension and the weaker the bearing structure stability and error tolerance. Step B13, Adaptive Matching of Process Weight Coefficients: Based on the characteristics of the multi-stage forming process of bearings, assign weight coefficients corresponding to the three types of scores. , , The weight normalization constraint is satisfied: In the formula, The raw material crystal phase weighting coefficient is the highest weighting coefficient because the characteristics of the raw material substrate determine the strength of the bearing basic structure. The tooling clamping stiffness weighting coefficient affects the consistency of grinding and ultra-precision machining processes. This is a weighting coefficient for machine tool machining accuracy, reflecting the characteristics of instantaneous machining error fluctuations. This weighting allocation aligns with the priority of bearing manufacturing defect causes, differing from the common calculation method of equal weighting, and improving the process adaptability of the endowment index calculation. Step B14, Weighted Solution of Individual Process Endowment Index: Substitute the standardized scoring parameters and adaptive weight parameters into the endowment index calculation formula to accurately solve the individual process endowment index of a single bearing. The formula is as follows: ; Step B15, Endowment Index Validation and Parameter Alignment: Verify the validity of the obtained endowment index. Perform range validation to remove invalid calculation results that exceed the range; simultaneously, include compliant results... The parameters are cached and the total coupling hazard of the single bearing is transmitted in step A4. This creates a one-to-one matching parameter set, which is then substituted into the dynamic threshold correction formula in the next step. Complete adaptive threshold correction.

[0014] Preferably, the specific implementation steps for the adaptive quality inspection threshold dynamic correction in step B2 are as follows: Step B21, Core Correction Parameter Collection and Validity Verification: Collect the industry standard fixed qualified thresholds used in this threshold calculation. (Based on the benchmark threshold set according to the national standard production and quality inspection specifications for the corresponding bearing model), fixed threshold correction factor (A preset constant adjustment coefficient is used to control the threshold correction range and avoid excessive correction that deviates from industry standards), Individual bearing process endowment index. (Step B1 is solved, with values ​​[0,1] representing the tolerance of individual errors), total coupling hazard of multi-process errors. (A value ≥ 0 indicates the risk level of latent defects in the bearing). All parameters undergo range validation to eliminate out-of-bounds data and ensure the accuracy of threshold correction. Step B22, Calculation using the dynamic threshold correction formula: Substitute all compliant parameters into the dynamic threshold correction formula to solve for the personalized adaptive quality inspection threshold for a single bearing. The core formula is as follows: ;in The corrected individual adaptive quality inspection threshold; Set a fixed pass threshold for industry standards; This is the threshold correction coefficient, fixed at 0.3, to control the threshold correction magnitude and avoid excessive deviation from the standard; the formula is... To differentiate the quality of individual bearings, a threshold of 0.5 is used to distinguish between high-tolerance and low-tolerance bearings; through... The linkage of the residual error coupling decoupling feature extraction module in multiple processes enables bidirectional coupling correction of individual endowment characteristics and implicit coupling risks. Unlike the simple algorithm that relies on a single parameter for correction, the correction logic is fully aligned with the individual differences and defect causes in the multi-process production of bearings. Step B23, Implementation of scenario-based adaptive threshold correction logic: Based on step 22 The calculation results, combined with the actual production characteristics of bearings, yield four precisely tailored threshold correction conditions to specifically address over-inspection and under-inspection issues: Operating Condition 1: High-Endowment, Low-Risk, High-Quality Bearings: When and At that time, the bearing raw materials were uniform, the processing stability was high, and there were no hidden defects caused by the coupling of errors from multiple processes. After the formula calculation... Expand the quality inspection threshold; this type of bearing has strong tolerance to error, and slight process errors will not cause service failure. Relaxing the threshold can prevent high-quality bearings from being misjudged as defective products, reduce the waste of production resources, and solve the problem of over-inspection with the traditional unified threshold. Operating Condition 2: High-Endowment, High-Risk Coupled Bearings: When and At that time, the bearing has excellent basic tolerance, but there is a significant hidden danger of multi-process error coupling and superposition. After the formula calculation, the threshold dropped slightly, the judgment standard was appropriately tightened, the hidden coupling defects were accurately identified, and the inherent advantages were avoided from masking the risk of superposition failure. Operating Condition 3: Low-endowment, low-risk general bearings: when and At that time, the individual structural stability and error tolerance of the bearing were relatively weak, but there were no hidden coupling defects. After formula calculation... Appropriately tighten the threshold, strictly control the basic machining accuracy of low-endowment bearings, and avoid the risk of failure in later service life in advance. Operating Condition 4: Low-endowment, high-risk, weak bearings: when and At the same time, bearings simultaneously suffer from individual process defects and multi-process error coupling risks, resulting in a double failure risk. After formula calculation, the threshold is significantly tightened, and the strict quality inspection standards are used to accurately intercept high-risk defective products, solving the problem of missed detection by the traditional unified threshold. Step B24, Threshold Boundary Constraints and Compliance Calibration: To avoid excessive deviation of the dynamically corrected threshold from industry standards and to ensure the compliance of quality inspection results, the personalized threshold obtained is adjusted accordingly. Apply boundary constraints to limit the corrected threshold range. ; Step B25, Personalized Threshold Fixing and Parameter Binding: Apply the calibrated single-bearing-specific adaptive quality inspection threshold... With the current bearing , Each parameter is bound and stored in a fixed manner, forming a unique quality inspection standard for each bearing, providing a unique and accurate quantitative basis for the next step of differentiated quality level determination. Compared with the prior art, the beneficial effects of the present invention are: This invention significantly reduces the risk of early bearing failure by accurately identifying coupled hidden defects. Employing a layered coupling-decoupling algorithm, it separates the hazards of single-process errors from the cumulative effects of multiple processes. This allows for the detection of bearings that meet the standards for each process but whose errors are easily coupled and could lead to abnormal noise and fracture, thus overcoming the shortcomings of traditional single-process inspection.

[0015] This invention features adaptive differentiated quality inspection, balancing production capacity and product quality. Based on an individual process endowment index, the judgment threshold is dynamically adjusted. Standards are appropriately relaxed for bearings with excellent base materials and high durability to avoid misjudging and scrapping high-quality parts; while thresholds are tightened for bearings with weak endowments to intercept potential defects and reduce resource waste.

[0016] The system of this invention has self-iterative optimization capabilities. The process coupling correction coefficient is updated in reverse with each batch of quality inspection data to adapt to fluctuations in raw materials and machine tool loads. Under long-term operation, the rate of missed inspections and false judgments continues to decrease, adapting to various production conditions.

[0017] This invention enables the quantification and traceability of defect causes. It outputs quantitative parameters such as independent hazards of processes, cross-coupled hazards, and process endowments, which can accurately distinguish whether quality problems originate from single-process deviations or the cumulative effects of multiple processes, guiding the optimization of processes such as forging and grinding. Attached Figure Description

[0018] Figure 1 This is a schematic diagram of the system structure of the present invention; Figure 2 This is a schematic diagram of the workflow of the multi-process residual error coupling and decoupling feature extraction module of the present invention; Figure 3 This is a schematic diagram of the workflow of the individual process endowment adaptive threshold quality inspection judgment module of the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] Please see Figure 1-3 This invention provides a technical solution: a bearing production quality inspection system based on production data analysis, comprising a data acquisition module, a multi-process residual error coupling and decoupling feature extraction module, an individual process endowment adaptive threshold quality inspection judgment module, a data storage module, and a result output module. The data acquisition module collects two types of data sources: the first type is multi-process forming residual error data, including four types of original error parameters: forging process density residual deviation, heat treatment process stress residual deformation, grinding process dimensional residual deviation, and ultra-precision process roughness residual deviation; the second type is individual process endowment microscopic data, including three types of original process parameter data: raw material crystal phase uniformity, tooling clamping pressure fluctuation value, and machine tool instantaneous machining accuracy. The collected data is then cleaned, normalized, and standardized to obtain preprocessed data. The specific implementation steps of this module are as follows: Step 1: Precise Classification and Parameter Definition of Data Collection Objects: Two types of data sources are collected: The first type is multi-process forming residual error data, which is adapted to the coupling and decoupling calculation requirements of the multi-process residual error coupling and decoupling feature extraction module. It includes four types of original error parameters: forging process density residual deviation, heat treatment process stress residual deformation, grinding process dimensional residual deviation, and ultra-precision process roughness residual deviation. The second type is individual process endowment micro data, which is adapted to the endowment index calculation requirements of the individual process endowment adaptive threshold quality inspection judgment module. It includes three types of original process parameters: raw material crystal phase uniformity, tooling clamping pressure fluctuation value, and machine tool instantaneous machining accuracy, which accurately correspond to the three core causes of individual differences in bearings of the same batch. Step 2: Synchronous Raw Data Acquisition Across the Entire Process: Utilizing high-precision sensing and detection equipment, machine tool operation data acquisition terminals, and material testing equipment on the production line, data for each bearing across the entire production chain is collected synchronously on a one-to-one basis. This achieves independent data collection for individual bearings and batch data collection for all batches. Raw error data is collected for each of the four core processing steps. For the entire process from raw material warehousing, machining clamping, to instantaneous cutting, microscopic process parameters are collected synchronously, ensuring that the error data for each bearing is bound to its process inherent data, forming a dedicated raw dataset for each bearing and avoiding data misalignment and batch confusion.

[0021] Step 3: Anomaly Filtering and Cleaning of Multi-Source Raw Data: All collected raw data undergoes anomaly screening and removal, eliminating two types of invalid data: first, extreme anomaly data caused by equipment malfunctions, signal interference, or human error; second, incomplete sample data due to missing data or data acquisition interruptions. Only valid raw data that truly reflects the bearing production status is retained, preventing anomaly data from interfering with the accuracy of subsequent algorithm calculations and ensuring the validity and reliability of data throughout the entire process.

[0022] Step 4: Heterogeneous Data Normalization: Due to the different dimensions and vastly different numerical ranges of the collected raw data, they cannot be directly substituted into the algorithm. Therefore, an extreme value normalization algorithm is used to uniformly standardize all the raw data, mapping all parameters to... The standard range, adapted to the parameter value specifications of this invention's algorithm, has the following core normalization calculation formula: ;in The standardized parameters have a strictly defined value range. ; For various types of raw data; , These are the maximum and minimum values ​​of the parameters corresponding to the current production batch; normalization is performed based on the batch's full-domain data to adapt to the batch's production conditions and improve parameter adaptability.

[0023] For the process error data, the standardized residual error of the four processes is obtained after standardization using the above formula. , , , It is directly used in the multi-process residual error coupling and decoupling feature extraction module to calculate covariance, variance, and coupling hazard; for individual process parameters, it is standardized to obtain raw material scores. Clamping stability score Machine tool accuracy rating The process endowment index is directly used in the individual process endowment adaptive threshold quality inspection judgment module. Solve to achieve a unified format and standard for all parameters.

[0024] Step 5: Standardize and Classify Structured Data Encapsulation. Based on the differentiated input requirements of the multi-process residual error coupling and decoupling feature extraction module and the individual process endowment adaptive threshold quality inspection judgment module, the standardized parameters are classified and encapsulated to construct two types of dedicated structured datasets: one is a multi-process error dataset, containing... , , , The first is an algorithm for adapting the multi-process residual error coupling and decoupling feature extraction module to a hierarchical coupling and decoupling algorithm; the second is an individual process endowment dataset, containing... , , It features an adaptive threshold quality inspection module with dynamic threshold correction algorithm tailored to individual process characteristics. All data retains a unique identifier for each bearing, achieving precise binding of "one data point per bearing".

[0025] Step 6, Targeted Data Distribution and Pre-caching: The encapsulated structured data is precisely distributed to the corresponding modules: The multi-process standardized residual error dataset is transmitted in real-time to the multi-process residual error coupling and decoupling feature extraction module, supporting process correlation statistics, coupling coefficient calculation, and coupling hazard determination; the individual process standardized score dataset is transmitted in real-time to the individual process endowment adaptive threshold quality inspection judgment module, supporting endowment index calculation. Simultaneously, all standardized data is cached in the system database, providing raw data support for subsequent batch iterative optimization, process traceability, and parameter iterative updates, forming a data closed loop.

[0026] The multi-process residual error coupling and decoupling feature extraction module receives preprocessed data from the acquisition module and aggregates residual error data from multiple processes including bearing forging, heat treatment, grinding, and ultra-precision machining. Through a multi-dimensional residual error hierarchical coupling and decoupling algorithm, it achieves hierarchical decoupling of independent errors in single processes and coupled errors across multiple processes, quantifies the overall hazard of error coupling, and extracts latent defect features. The specific implementation steps are as follows: Step A1: Collection and Standardization of Residual Error Data from Multiple Processes: Receive preprocessed data from the acquisition module and collect four categories of original residual error data: forging density deviation of a single bearing, residual deformation from heat treatment, grinding dimensional deviation, and ultra-precision roughness deviation. Remove abnormal extreme values ​​caused by equipment failures. Use an extreme value normalization method to uniformly transform process error data of different dimensions and ranges into standardized residual errors within the 0~1 range. Eliminate the impact of dimensional differences on coupling analysis; Step A2, Statistical Analysis of Process Error Correlation and Coupling Coefficient Calculation: Based on the multi-process error dataset of bearings in the same batch, calculate the covariance and variance of each pair of process errors, substitute them into the coupling coefficient calculation formula, and obtain the error coupling influence coefficient between any two processes. Combined with bearing model matching coupling correction coefficient The coupling coefficient is adaptively calibrated to accurately match the forming characteristics of different bearings; the specific implementation steps are as follows: Step A21: Construction of multi-process error pairing dataset: Standardize residual errors based on the four core processes of a single bearing. forging, Heat treatment Grinding, Ultra-precision is the basic unit. It collects all sample data from the current complete production batch, constructs a pairwise paired error dataset for each process, and forms... , , , , , There are six sets of process pairings, covering all cross-process error coupling scenarios, providing complete data support for subsequent correlation calculations; Step A22, Solving for statistical characteristic quantities of process error: For each pair of processes... Calculate the covariance of process errors and the variance of single-process errors separately, specifically by setting up statistical rules: solve for the covariance of the two sets of process errors. , used to characterize the , The degree of correlation between residual errors in two processes is considered; a larger value indicates a stronger coupling and correlation between the errors of the two processes. The variance of single-process errors is calculated simultaneously. , It is used to characterize the degree of fluctuation and dispersion of residual error in a single process in batch production, and to reflect the stability of single-process processing; Step A23, Initial Coupling Correlation Coefficient Calculation: Substitute the covariance and variance data of each process group into the uncalibrated basic coupling coefficient formula to calculate the initial coupling correlation coefficient: ;in The initial coupling coefficient, which is not adapted to the bearing process characteristics, only reflects the correlation of process error from the perspective of data statistics, without distinguishing the differential coupling hazards of different bearing structures and processing characteristics; Step A24, Bearing Model Adaptive Coefficient Calibration: Based on the bearing types of the current production line, match the process coupling correction coefficient. Complete adaptive calibration; process coupling correction coefficient The specific calculation logic is as follows: Step A24.1, Calculation of Discrete Entropy of Single-Process Error: To quantify the degree of process fluctuation disorder in the current batch production of each process, a discrete entropy model of process error is constructed to characterize the disordered fluctuation characteristics of residual error in a single process. The formula for calculating the discrete entropy of single-process error is as follows: , specific For the first The discrete entropy of the error of each process, where the process number is... These correspond to forging, heat treatment, grinding, and ultra-precision processes, respectively. For the first Standardization residual error of each process The probability distribution density across all samples in the current batch; The number of intervals for process error statistics; The larger the value, the more drastic the process fluctuations, and the higher the risk of small residual errors in the process coupling and superimposing to form hidden defects. Step A24.2, Batch Global Process Correlation Calculation: Based on the inherent process sequence logic of the four bearing processes, a batch global process correlation coefficient is constructed to quantify the overall linkage coupling strength of the multi-process errors in the entire batch. The calculation formula is as follows: ;in This is the batch-wide process correlation coefficient, with a value range of [value range missing]. The larger the value, the stronger the correlation between multiple processing steps and the more significant the superposition and coupling effect of process errors. Step A24.3, Dynamic Coupling Correction Coefficient Fusion Solution: Combining the discrete entropy of single-process fluctuation characteristics with the correlation degree of global process linkage characteristics, an adaptive fusion algorithm is constructed to dynamically solve the coupling correction coefficient adapted to the current batch production conditions. Core calculation formula: ;in The mean of the discrete entropy of the errors in the four processes satisfies This is used to characterize the overall process fluctuation level of the current batch of bearings; a constant of 1.0 serves as the baseline correction base to ensure the stability of the parameter baseline; [This is to achieve...] The value is dynamically and adaptively adjusted according to the operating conditions, and eventually stabilizes at a certain value. interval; Step A24.4, Adaptive Logic for Operating Conditions: When producing precision miniature bearings, experiencing drastic batch process fluctuations, and having strong process interrelationships, and Increase synchronously. Automatically approaches 1.2, appropriately amplifies the weight of process coupling errors, and accurately captures minute hidden coupling defects; when producing heavy-duty industrial bearings, with high batch process stability and weak process linkage, and Decrease Automatically approaches 0.9, weakens invalid slight coupling interference, avoids misjudgment caused by over-detection, and achieves adaptive matching under all working conditions.

[0027] Step A25, Final Coupling Influence Coefficient Solution and Consolidation: The initial coupling coefficients are then... Category Adjustment Factor Substitute into the coupling coefficient formula The final error coupling influence coefficients corresponding to the six pairs of process operations were calculated one by one. , The actual coupling hazard weights after the superposition of errors from different processes were quantified, and adaptive calibration of all process coupling parameters was completed. The solved parameters were then used to calculate the total... The parameters are output to step A3.

[0028] Step A3, Layered Decoupling and Coupling Hazard Calculation: Using the coupling decoupling formula, the independent error impact value of a single process and the cross-coupling error impact value of two processes are calculated layer by layer, and finally the total coupling hazard of a single bearing is obtained by summing them up. ; It can accurately quantify the hidden defect risks resulting from the superposition of compliance errors in multiple processes, making up for the shortcomings of traditional single-process inspection in that it cannot identify coupled defects; the specific implementation steps are as follows: Step A31, Core Calculation Parameter Collection and Validity Verification: First, collect all the prerequisite parameters required for this step, complete the parameter consistency verification, and ensure calculation accuracy: standardize residual errors in four processes. , , , (Values ​​range from 0 to 1), Independent error influence coefficient for each process The dynamic coupling influence coefficient of the six process pairings (Adaptively corrected by discrete entropy and global correlation) (Obtained); invalid samples with abnormal parameters or missing data are removed to ensure that all input parameters are valid data for adaptive solution of the current bearing batch operating conditions; Step A32, Calculation of Independent Hazardous Component of Single Process Error: Based on the hierarchical decoupling concept, the independent hazard effect of residual error in a single process is first isolated. This component represents the fundamental hazard of a small residual error in a single process to bearing quality, without cross-process superposition effect. The calculation formula is as follows: ;in, This represents the total hazard component of independent errors in a single process. The preset single-process independent error influence coefficient is used to characterize the basic hazard weight of errors in forging, heat treatment, grinding, and ultra-precision single processes; This step is for the standardized residual error of the corresponding process. It can independently quantify the basic negative impact of each qualified residual error, solving the problem that traditional technology can only determine the error is qualified but cannot quantify the basic degree of harm. Step A33, Calculation of the Hazardous Component of Cross-Process Coupling Errors: Calculate the additional hazard component caused by the superposition of errors across processes. The calculation formula is as follows: In the formula, This refers to the component of errors caused by cross-coupling of multiple processes; The dynamic process coupling influence coefficient, which is adaptively solved in step A2, integrates the entropy value of the current batch process fluctuation and the process correlation characteristics, and can accurately adapt to real-time production conditions. This is the coupled superposition of residual errors from two processes, characterizing the synergistic effect of minute errors in different processes. This component specifically quantifies the hidden defects such as stress concentration and contact misalignment caused by the coupling of errors in multiple processes; Step A34: Calculation of Total Coupling Hazard: Combining the basic hazard components of a single process with the superimposed hazard components of multiple processes, the total coupling hazard is obtained, fully characterizing the overall defect risk of the bearing. Substituting this into the coupling decoupling formula: ;in The total coupling hazard of residual errors from multiple processes in the bearing is represented, and its value is greater than or equal to 0. The numerical classification logic is as follows: The closer it is to 0, the more uniform the errors in each process of the bearing are, the weaker the coupling and superposition effect is, and the less risk there is of hidden service defects. The higher the value, the more significant the cumulative effect of compliance errors in multiple processes, and the higher the risk of latent bearing failure.

[0029] Step A35, Latent Defect Risk Quantification and Calibration: Based on the solution obtained... Complete defect risk stratification and labeling, which differs from the traditional coarse judgment logic of judging a product as good as that of a single process passing: single process error meets the standard but... Bearings exceeding the adaptive threshold are identified as risk samples with latent coupling defects, enabling the identification of latent defects that traditional testing methods cannot cover, thus overcoming the shortcomings of existing quality inspection technologies in detecting these defects. Step A4, Feature Output and Data Transmission: The calculated total coupling hazard degree... The independent and cross-coupled error characteristic data of each process are packaged and output to the individual process endowment adaptive threshold quality inspection judgment module, providing accurate quantitative basis for individual process endowment threshold correction and differentiated quality judgment. The individual process endowment adaptive threshold quality inspection judgment module, in conjunction with the multi-process residual error coupling and decoupling feature extraction module, uses data linkage. Based on the quantified error coupling hazard degree and combined with the individual bearing micro-process parameters in the preprocessed data, it achieves personalized quality inspection threshold correction and differentiated quality judgment for a single bearing through a dynamic threshold correction algorithm based on process endowment differences. The specific implementation steps are as follows: Step B1, Individual Process Parameter Collection and Endowment Index Calculation: Collect microscopic process data from the entire production process of a single bearing: raw material crystal phase uniformity test data, machining tool clamping pressure fluctuation data, and machine tool instantaneous machining accuracy data. After standardizing and scoring these three types of data, substitute them into the endowment index calculation formula to obtain the process endowment index specific to each single bearing. This accurately characterizes an individual's tolerance to error; the specific implementation steps are as follows: Step B11: Precise Collection and Anomaly Filtering of Multi-Dimensional Microscopic Process Parameters: Three types of core original process data that cause individual bearing endowment differences within the same batch are specifically collected: 1) Original test data on the uniformity of raw material crystal phase, characterizing the structural uniformity and stability of the bearing substrate; 2) Data on the fluctuation of clamping pressure of machining fixtures, characterizing the machining stability deviation during assembly and clamping; 3) Data on the instantaneous machining accuracy of the machine tool, characterizing the fluctuation of equipment operating accuracy during the machining of a single bearing. Simultaneously, the data cleaning logic of the multi-process residual error coupling and decoupling feature extraction module is used to eliminate extreme abnormal data caused by equipment failure and human error, retaining effective sample data that truly reflects the production characteristics of a single bearing, ensuring the accuracy of subsequent endowment calculations. Step B12, Multidimensional Parameter Normalization and Standardization Scoring: Due to the inconsistency in the dimensions and numerical ranges of the three types of original process parameters, direct weighted calculation is not possible. Therefore, the original data is normalized to convert them into a unified standard. The standardized scores for each interval yielded the following: raw material phase uniformity score. Tooling clamping stability score Machine tool machining accuracy rating The scoring rules are uniformly adapted to the judgment logic of this invention: the closer the score value is to 1, the better the process performance in that dimension and the stronger the bearing error tolerance; the closer the score value is to 0, the more minor defects exist in the process in that dimension and the weaker the bearing structure stability and error tolerance. Step B13, Adaptive Matching of Process Weight Coefficients: Based on the characteristics of the multi-stage forming process of bearings, assign weight coefficients corresponding to the three types of scores. , , The weight normalization constraint is satisfied: In the formula, The raw material crystal phase weighting coefficient is the highest weighting coefficient because the characteristics of the raw material substrate determine the strength of the bearing basic structure. The tooling clamping stiffness weighting coefficient affects the consistency of grinding and ultra-precision machining processes. This is a weighting coefficient for machine tool machining accuracy, reflecting the characteristics of instantaneous machining error fluctuations. This weighting allocation aligns with the priority of bearing manufacturing defect causes, differing from the common calculation method of equal weighting, and improving the process adaptability of the endowment index calculation. Step B14, Weighted Solution of Individual Process Endowment Index: Substitute the standardized scoring parameters and adaptive weight parameters into the endowment index calculation formula to accurately solve the individual process endowment index of a single bearing. The formula is as follows: ; Step B15, Endowment Index Validation and Parameter Alignment: Verify the validity of the obtained endowment index. Perform range validation to remove invalid calculation results that exceed the range; simultaneously, include compliant results... The parameters are cached and the total coupling hazard of the single bearing is transmitted in step A4. This creates a one-to-one matching parameter set, which is then substituted into the dynamic threshold correction formula in the next step. Complete adaptive threshold correction Step B2, Adaptive Quality Inspection Threshold Dynamic Correction: Retrieve Industry Standard Fixed Thresholds The coupling hazard degree transmitted by the multi-process residual error coupling decoupling feature extraction module is combined with the feature extraction module. The endowment index calculated in step B1 The personalized quality inspection threshold for a single bearing is calculated using a dynamic threshold correction formula. Specific correction logic: For high-quality bearings with high endowment index and low coupling hazard, the threshold is appropriately relaxed to avoid over-inspection due to minor errors; for weak bearings with low endowment index and high coupling hazard, the threshold is appropriately tightened to intercept potential defective products in advance; the specific implementation steps are as follows: Step B21, Core Correction Parameter Collection and Validity Verification: Collect the industry standard fixed qualified thresholds used in this threshold calculation. (Based on the benchmark threshold set according to the national standard production and quality inspection specifications for the corresponding bearing model), fixed threshold correction factor (A preset constant adjustment coefficient is used to control the threshold correction range and avoid excessive correction that deviates from industry standards), Individual bearing process endowment index. (Step B1 is solved, with values ​​[0,1] representing the tolerance of individual errors), total coupling hazard of multi-process errors. (A value ≥ 0 indicates the risk level of latent defects in the bearing). All parameters undergo range validation to eliminate out-of-bounds data and ensure the accuracy of threshold correction. Step B22, Calculation using the dynamic threshold correction formula: Substitute all compliant parameters into the dynamic threshold correction formula to solve for the personalized adaptive quality inspection threshold for a single bearing. The core formula is as follows: ;in The corrected individual adaptive quality inspection threshold; Set a fixed pass threshold for industry standards; This is the threshold correction coefficient, fixed at 0.3, to control the threshold correction magnitude and avoid excessive deviation from the standard; the formula is... To differentiate the quality of individual bearings, a threshold of 0.5 is used to distinguish between high-tolerance and low-tolerance bearings; through... The linkage of the residual error coupling decoupling feature extraction module in multiple processes enables bidirectional coupling correction of individual endowment characteristics and implicit coupling risks. Unlike the simple algorithm that relies on a single parameter for correction, the correction logic is fully aligned with the individual differences and defect causes in the multi-process production of bearings. Step B23, Implementation of scenario-based adaptive threshold correction logic: Based on step 22 The calculation results, combined with the actual production characteristics of bearings, yield four precisely tailored threshold correction conditions to specifically address over-inspection and under-inspection issues: Operating Condition 1: High-Endowment, Low-Risk, High-Quality Bearings: When and At that time, the bearing raw materials were uniform, the processing stability was high, and there were no hidden defects caused by the coupling of errors from multiple processes. After the formula calculation... Expand the quality inspection threshold; this type of bearing has strong tolerance to error, and slight process errors will not cause service failure. Relaxing the threshold can prevent high-quality bearings from being misjudged as defective products, reduce the waste of production resources, and solve the problem of over-inspection with the traditional unified threshold. Operating Condition 2: High-Endowment, High-Risk Coupled Bearings: When and At that time, the bearing has excellent basic tolerance, but there is a significant hidden danger of multi-process error coupling and superposition. After the formula calculation, the threshold dropped slightly, the judgment standard was appropriately tightened, the hidden coupling defects were accurately identified, and the inherent advantages were avoided from masking the risk of superposition failure. Operating Condition 3: Low-endowment, low-risk general bearings: when and At that time, the individual structural stability and error tolerance of the bearing were relatively weak, but there were no hidden coupling defects. After formula calculation... Appropriately tighten the threshold, strictly control the basic machining accuracy of low-endowment bearings, and avoid the risk of failure in later service life in advance. Operating Condition 4: Low-endowment, high-risk, weak bearings: when and At the same time, bearings simultaneously suffer from individual process defects and multi-process error coupling risks, resulting in a double failure risk. After formula calculation, the threshold is significantly tightened, and the strict quality inspection standards are used to accurately intercept high-risk defective products, solving the problem of missed detection by the traditional unified threshold. Step B24, Threshold Boundary Constraints and Compliance Calibration: To avoid excessive deviation of the dynamically corrected threshold from industry standards and to ensure the compliance of quality inspection results, the personalized threshold obtained is adjusted accordingly. Apply boundary constraints to limit the corrected threshold range. ; Step B25, Personalized Threshold Fixing and Parameter Binding: Apply the calibrated single-bearing-specific adaptive quality inspection threshold... With the current bearing , The parameters are bound and stored in a fixed manner to form a unique quality inspection and judgment standard for each bearing, providing a unique and accurate quantitative basis for the next step of differentiated quality level judgment.

[0030] Step B3, Differentiated Quality Grade Determination: The actual process error and coupling error of a single bearing are compared with the personalized correction threshold. A four-level judgment standard is preset: Excellent (error far below the correction threshold), Qualified (error meets the correction threshold requirements), Warning (error close to the correction threshold, posing a potential risk), and Unqualified (error exceeds the correction threshold). Compared to traditional uniform threshold judgment, this achieves precise differentiated quality inspection. The specific implementation steps are as follows: Step B31, Collection and Establishment of Core Parameters: All prerequisite core parameters for this grading determination are collected uniformly to ensure that the determination data is completely consistent with the algorithm calculation results and has no deviation: Single bearing adaptive correction of quality inspection threshold. Industry fixed standard threshold Standardized residual errors in each process , , , Total Hazard of Multi-Process Error Coupling .by As the sole core benchmark for judging the quality of a single bearing, it replaces the traditional unified fixed threshold. The stratified determination is completed by combining individual endowment differences with coupling defect risks; Step B32, Adaptive Division of the Fourth-Level Judgment Interval: Based on the fault tolerance requirements of industrial bearing production quality inspection, the threshold is adjusted in a personalized manner. Based on this, a unified interval ratio coefficient is introduced to divide the four independent judgment intervals. The interval division is applicable to all bearing specifications. The specific interval division rules are as follows: a safety margin coefficient of 0.8 and a risk warning coefficient of 0.95 are set to define the boundaries of the four quality levels, so as to achieve quantitative and standardized differentiated judgment.

[0031] Step B33, Layered Precision Quality Grade Determination: Combining the overall level of actual error of a single bearing with the total coupled hazard level. Matching the corresponding judgment interval, the four-level quality level is accurately determined. The judgment standards and technical principles for each level are as follows: Step B33.1, Quality Product Judgment: When the maximum value of the bearing's comprehensive error... And the total coupling hazard At that time, it was judged to be a high-quality product. The individual process endowment index of this type of bearing... High-speed, multi-process error has no coupling and superposition effect, excellent processing accuracy, no hidden service defects, and sufficient error tolerance margin, which can be adapted to high-speed, heavy-load, and long-life high-end service conditions. Step B33.2, Quality Inspection: When the maximum value of the bearing's overall error is within... Interval, and total coupling hazard degree When the bearings are within the normal and reasonable range, they are judged to be qualified products. The residual errors in the manufacturing process of this type of bearing are controllable, the risk of coupled superposition is extremely low, the individual process stability meets the requirements of normal service, there is no risk of early failure, and they meet the batch production standards. Compared with the traditional unified threshold judgment, this invention can accurately retain bearings with excellent endowment, slight errors but meeting performance standards, avoiding the scrapping of high-quality products due to over-inspection. Step B33.3, Warning Item Judgment: When the maximum value of the bearing's overall error is within... Interval, or total coupling hazard When the levels are too high or close to the risk threshold, the product is classified as a warning product. Although this type of bearing does not exceed the individual quality inspection threshold, the process error is close to the critical value, and there is a slight risk of multiple process error coupling and superposition. Combined with the characteristic of weak individual process endowment, there is a potential risk of wear and abnormal noise in long-term service. It needs to be sorted and re-inspected and rated separately, and is prohibited from being shipped as a high-end qualified product. Step B33.4, Determination of Non-conforming Products: When the maximum value of the bearing's overall error... or total coupling hazard Products exceeding the risk threshold or exhibiting significant latent defects are deemed non-conforming. These bearings or processes exhibit excessive machining errors, or multiple processes exhibit coupled and cumulative errors leading to severe latent defects. Even if a single process error meets industry standards, limitations imposed by individual process capabilities can easily result in early service failure. This method precisely intercepts high-risk defective products that traditional quality inspections might miss.

[0032] Step B34, Grade Labeling and Risk Traceability Marking: For each bearing that has completed the Level 4 assessment, a unique quality label and risk traceability information are simultaneously bound: high-quality products are labeled with high adaptability to service conditions; warning products and non-conforming products are simultaneously associated with corresponding process errors. Coupling Hazard Technology endowment index Parameters, accurately pinpointing the quality risks caused by excessive errors in a single process or the coupling and superposition of errors in multiple processes, provide data support for subsequent process optimization and defect traceability; Step B35, Compliance Verification of Judgment Results: Perform batch verification of the quality grade judgment results for all bearings, relying on the threshold boundary constraint rules in Step B2 to ensure that all judgment criteria are met. All in Within the compliance range, we will eliminate the distortion of judgments caused by extreme thresholds and ensure that differentiated quality inspection results are both technologically innovative and in line with industry factory quality inspection standards. Step B4, Data Reverse Iteration and Result Output: Output the quality inspection judgment result of a single bearing, and simultaneously store the individual bearing endowment parameters, threshold correction data, and coupling error matching data. Iterate backwards to optimize the process coupling correction coefficient of the multi-process residual error coupling decoupling feature extraction module. This improves the decoupling accuracy of coupling errors in subsequent batches, enabling continuous system optimization. The specific implementation steps are as follows: Step B41, Collection and Compliance Screening of Core Data for End-to-End Quality Inspection: Unify the collection of calculation and judgment data for the entire process of a single bearing to form a complete individual quality inspection dataset. Core parameters include: standardized residual error of each process. , , , Total Coupling Hazard Individual craft endowment index Industry benchmark threshold Personalized correction threshold Final quality assessment level, batch original process fluctuation discrete entropy Global process correlation Original coupling correction coefficients before iteration ; Step B42, Batch Iterative Sample Set Construction and Error Deviation Quantification: Data from a single bearing is used only as single-sample data. After all bearings in the current production batch have completed quality inspection, all valid sample data are collected to construct a batch iterative sample set. Based on the actual judgment results of the sample set, the algorithm matches the deviation and constructs an iterative correction core deviation factor. Used to characterize the current The algorithm judgment error corresponding to the parameters is calculated using the following formula: ;in This represents the average batch judgment deviation; a larger value indicates a higher current coupling correction coefficient. The worse the compatibility with the working conditions of this batch; The total number of valid samples in the batch; , , The first The individual threshold, process endowment index, and total coupling hazard of each bearing; the deviation factor is generated entirely based on the actual quality inspection data of this batch.

[0033] Step B43, Coupling Correction Coefficient Reverse adaptive iterative update. Based on the feature extraction module for decoupling residual errors from multiple processes. The basic solution model, combined with batch deviation factor Construct a dynamic iterative update formula and correct the original formula. The parameters are adapted to the actual production conditions of the current batch, and the iterative update formula is as follows: ;in The coupling correction coefficient is used for the newly updated process. The learning rate is fixed at 0.15 and is used to control the magnitude of correction in a single iteration, so as to avoid algorithm oscillation caused by sudden parameter changes. The average deviation for batch determination; , The mean discrete entropy of the current batch of processes and the global process correlation degree are solved by the feature extraction module for decoupling the coupling of residual errors in multiple processes. The parameters of the previous module are fully reused to ensure that the iterative logic is from the same source and closed loop.

[0034] The core logic of this iterative mechanism is: when the current... Poor adaptability and judgment bias If the value is too high, it will automatically make a small correction. The numerical values ​​optimize the calculation accuracy of the process coupling weights, precisely adapting to the characteristics of the raw materials, machine tool conditions, and process fluctuations of this batch.

[0035] Step B44, Parameter Boundary Calibration and Algorithm Model Consolidation: To ensure the parameters are compliant and effective after iteration, the new coupling correction coefficients are adjusted. Perform interval constraint calibration, continuing the compliant interval rules mentioned above, and limit... After calibration The coupling coefficient calculation model of the multi-process residual error coupling and decoupling feature extraction module is updated synchronously to replace the original parameters before iteration. This completes the linkage upgrade of the algorithm models of the multi-process residual error coupling and decoupling feature extraction module and the individual process endowment adaptive threshold quality inspection judgment module. This provides more accurate core parameter support for the calculation of process coupling correlation degree and the solution of coupling hazard degree for the next batch of bearings, and continuously reduces the probability of missed detection and misjudgment of hidden defects.

[0036] Step B45, Multi-dimensional Standardized Result Output and Traceability Archiving: The final output is the standardized quality inspection result for a single bearing, including: individual bearing process endowment level, multi-process error coupling risk level, adaptive quality inspection threshold parameters, final quality judgment label, and process deviation traceability information. Simultaneously, all iterative data, updated algorithm parameters, and batch quality inspection reports are synchronously stored in the system database, forming a complete closed loop of "data acquisition - coupling / decoupling - differentiated judgment - iterative optimization." This enables the system to continuously adaptively optimize with batch production, significantly improving the accuracy and adaptability of long-term batch quality inspection.

[0037] The data storage module receives and stores data from the entire data chain, including the data acquisition module, the multi-process residual error coupling and decoupling feature extraction module, the individual process endowment adaptive threshold quality inspection judgment module, and the result output module. The specific implementation steps are as follows: Step C1: Real-time Classification and Reception of Multi-Source Data Across the Entire Process: All compliant and valid data generated across the entire working process of the system are collected in real time and strictly categorized into four main types according to data attributes, fully matching the data caliber of the algorithm system described above, eliminating data mixing and misalignment: The first type is raw collected data, including raw residual error data from the four processes, raw material crystal phase uniformity, tooling clamping pressure fluctuations, machine tool instantaneous machining accuracy, and other raw process parameters; the second type is standardized preprocessed data, including normalized process standardized residual errors. , , , Process standardization score , , The third category consists of intermediate computational parameters of the algorithm, including the discrete entropy of process errors. Global process correlation Dynamic coupling correction coefficient Process coupling influence coefficient Single process independent hazard component Cross-coupling hazard components Batch judgment deviation factor The fourth category consists of final judgment and iteration parameters, including the total coupling hazard. Individual craft endowment index Industry benchmark threshold Personalized correction threshold Iterative update of coupling correction coefficients Single bearing quality grade label; Step C2, Hierarchical and Partitioned Structured Independent Storage: Based on data usage and computational hierarchy, a partitioned storage architecture is constructed. The four types of collected data are hierarchically and independently archived to ensure accurate and efficient cross-module data retrieval: First, a raw data storage area retains unprocessed raw production data for subsequent process traceability and parameter verification; second, a standardized parameter storage area provides dedicated storage for algorithm computation. , , , The system includes four main areas: a standardized parameter area to provide basic input for real-time computation; a third area for storing intermediate parameters, which stores dynamic computational parameters during coupling decoupling, endowment calculation, and threshold correction processes, supporting real-time algorithm iteration and fine-tuning; and a fourth area for storing results and iteration parameters, which stores the final quality inspection results. The parameters are iteratively updated to cover and optimize the baseline parameters for the next batch of algorithms. All data is bound to a unique identifier for each bearing, achieving a structured storage model of "one bearing, one file; one batch, one group". Step C3: Solidifying and Retaining Core Algorithm Parameter Versions: For the core adaptive algorithm parameters of the system, establish a batch version snapshot solidification mechanism to avoid the loss of historical parameters and batch operation disorder caused by dynamic iteration. This includes establishing the global process correlation for each batch of production. Mean of discrete entropy of process error Coupling correction coefficient Iterative learning rate Threshold correction coefficient Fixed baseline parameters and dynamic operating condition parameters are automatically archived and saved after batch quality inspection, forming a batch-specific parameter snapshot. This mechanism can accurately distinguish the algorithm adaptation parameters under different raw material batches and different machine tool operating conditions, ensuring the traceability and reproducibility of quality inspection results for different batches. Step C4: Real-time Update of Dynamic Iterative Dataset Cache: To address the system's bidirectional adaptive iterative optimization requirements, a separate iterative data cache partition is created to collect matching deviation data and parameter correction data in real time after the quality inspection of a single bearing. This continuously accumulates data from all valid samples within the batch. , , The parameter group is the batch deviation factor. calculate, Iterative updates provide a complete sample dataset. Cached data retains only valid data from the current production batch. After batch quality inspection, the data is automatically archived, solidified, and the real-time cache is cleared to prevent historical data from interfering with the accuracy of the current batch's algorithm iterations and to ensure real-time adaptation to operating conditions.

[0038] Step C5, Cross-Module Precise Data Retrieval and Output Support: Establish standardized data retrieval interfaces to precisely match the data requirements of each functional module, achieving targeted data transmission and reuse; provide standardized error parameters upwards to the multi-process residual error coupling and decoupling feature extraction module. With historical iteration parameters ; Provides an adaptive threshold quality inspection judgment module for individual process endowments. Historical process endowment parameter samples; ultimately providing the results output module with full-dimensional quality inspection data and traceability parameters to support the generation of standardized quality inspection reports.

[0039] Step C6, Long-term Data Archiving and Process Traceability Support: After completing batch iterative optimization and result output, archive all original data, standardized parameters, algorithm calculation data, quality inspection results, and iterative update parameters for that batch. The data is uniformly archived into a long-term database to build a full lifecycle data archive for bearings, encompassing production, testing, and optimization. Based on historical stored parameters, the causes of latent defects in individual bearings, batch process fluctuation patterns, and algorithm parameter adaptation trends can be traced back, providing data support for production line process optimization, machine tool maintenance, and raw material selection, while simultaneously meeting industrial quality inspection traceability compliance requirements.

[0040] The results output module receives the final data from the data storage module and the individual process endowment adaptive threshold quality inspection judgment module. Based on the calculation results of the coupling decoupling algorithm and the dynamic threshold correction algorithm, it completes the standardized integration, hierarchical output, working condition traceability display, and batch optimization result push of bearing quality inspection data. The specific implementation steps are as follows: Step D1, Accurate Retrieval and Compliance Verification of End-to-End Data: Retrieve final compliance data for individual bearings and batches from the data storage module in layers, ensuring that the output data and algorithm calculation data are completely identical and have zero deviation. The core retrieved parameters cover four main categories: First, individual process parameters, including raw materials, clamping, and machine tool standardization scoring. , , With the technology endowment index Secondly, it involves coupling defect quantification parameters, including standardized residual errors from each process. Independent hazard component Coupling Hazard Components Total Coupling Hazard Thirdly, threshold determination parameters, including industry benchmark thresholds. Personalized correction threshold Threshold correction coefficient Fourth, iterative optimization and judgment result parameters, including batch deviation factors. Coupling correction coefficient after iteration Single bearing four-level quality grade label. Simultaneously verify the compliance of all parameter ranges, eliminate invalid data due to iteration anomalies or failed judgments, and ensure accurate and reliable output results.

[0041] Step D2: Structured Integration and Classification of Multidimensional Quality Inspection Results: Scattered algorithm parameters are integrated into three categories of directly applicable structured output results to meet the needs of industrial production applications: The first category is the individual quality inspection results of a single bearing, used for sorting and grading a single bearing; the second category is the source tracing results of hidden defects, used to locate the causes of coupled risk of errors; the third category is the batch algorithm optimization results, used for iterative upgrades of quality inspection parameters for the next batch of production. The three categories of results are independent of each other and interconnected, fully covering the needs of quality inspection, source tracing, and optimization in all scenarios. Step D3: Precise output of differentiated quality grades for single bearings, based on the four-level judgment criteria in Step B3, combined with... Personalized threshold, comprehensive error value and The system couples risk values ​​to accurately output the final quality grade of a single bearing, including four categories: excellent, qualified, warning, and unqualified. Simultaneously, it links core parameters to supplement the judgment criteria: excellent bearings are simultaneously marked with high endowment and low coupled risk characteristics; warning and unqualified bearings are simultaneously associated with corresponding process errors. Coupling Hazard It clarifies whether the quality risk is caused by excessive error in a single process or by the coupling and superposition of errors in multiple processes, solves the shortcomings of traditional quality inspection that lacks judgment basis and risk traceability, and provides direct quantitative support for accurate sorting on the production line; Step D4, Visualization Output of Latent Defects and Individual Talent Differences: This specifically outputs latent information that traditional techniques cannot reveal: Firstly, it outputs the risk level of multi-process error coupling and superposition, through... The numerical values ​​intuitively reflect the degree of hidden harm caused by compliance residual errors; secondly, they output individual process endowment ratings, through... Numerical analysis distinguishes differences in bearing error tolerance capabilities, intuitively demonstrating the underlying basis for differentiated threshold correction. This output clearly demonstrates the innovative advantages of this invention compared to traditional uniform threshold quality inspection, enabling visualization of latent defects and quantitative display of individual differences. Step D5: Outputting Batch Adaptive Iterative Optimization Results: Based on the system's closed-loop iterative optimization mechanism, after batch production quality inspection, the core parameters of the batch algorithm optimization are output, including the batch average judgment deviation. Coupling correction coefficients after iterative update Batch process discrete entropy mean Global process correlation The optimized parameters are synchronously pushed to the multi-process residual error coupling and decoupling feature extraction module, directly overwriting the original parameters. The parameters provide updated benchmark parameters for calculating the error coupling coefficient and solving the severity of the next batch of bearings, realizing the visualized implementation of continuous iterative upgrades in the system's detection accuracy; Step D6, Standardized Report Generation and Full Data Closed-Loop Archiving: Standardized quality inspection reports for individual bearings and batches are automatically generated. These reports integrate five core components: original process data, algorithm parameters, differential judgment results, hidden risk tracing information, and iterative optimization parameters, fully meeting industrial quality inspection compliance and traceability requirements. Simultaneously, all output results are synchronously transmitted back to the data storage module for permanent archiving, forming a complete closed loop of "data acquisition - coupling / decoupling - threshold correction - differential judgment - result output - iterative optimization." This continuously accumulates production quality inspection data, providing long-term data support for production line process optimization, equipment maintenance, and raw material quality upgrades.

[0042] This invention discloses a bearing production quality inspection system based on production data analysis, belonging to the field of bearing intelligent manufacturing quality inspection technology. Addressing two industry-specific technical problems in existing bearing production quality inspection systems—namely, the coupling and superposition of multi-process compliance residual errors leading to latent failures, and the distortion of uniform threshold judgments due to individual process variations within the same batch of bearings—this invention abandons the conventional approach of independent single-process inspection and uniform static threshold judgment. Instead, it improves the algorithm by matching the multi-process forming mechanism of bearings with the individual production differences. A multi-process residual error coupling decoupling feature extraction module and an individual process-endowment adaptive threshold quality inspection judgment module form a closed-loop detection link of "precise decoupling of coupling errors - precise judgment of differentiated thresholds." The former achieves decoupling and separation of minor compliant residual errors from multiple processes and identification of latent defects, while the latter dynamically corrects the quality inspection threshold based on the decoupled coupling error data and individual bearing process endowment parameters. This solves the problems of missed latent failures, over-inspection of high-quality bearings, and misjudgment of defective bearings in batch production. This invention precisely adapts to the production characteristics of bearing forging, heating, grinding and other multi-process composite molding, filling the technical gap in the industry for residual error coupling and individual endowment difference quality inspection. The detection accuracy and industrial adaptability are greatly improved, and it has extremely high value for mass production implementation.

[0043] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A bearing production quality inspection system based on production data analysis, characterized in that, The system includes a data acquisition module, a multi-process residual error coupling and decoupling feature extraction module, an individual process endowment adaptive threshold quality inspection judgment module, a data storage module, and a result output module. The data acquisition module collects two types of data sources. The first type is multi-process forming residual error data, including four types of original error parameters: forging process density residual deviation, heat treatment process stress residual deformation, grinding process dimensional residual deviation, and ultra-precision process roughness residual deviation. The second type is individual process endowment microscopic data, including three types of original process parameter data: raw material crystal phase uniformity, tooling clamping pressure fluctuation value, and machine tool instantaneous machining accuracy. The collected data is then cleaned, normalized, and standardized to obtain preprocessed data. The multi-process residual error coupling and decoupling feature extraction module is used to receive the preprocessed data from the acquisition module and collect the residual error data of bearing forging, heat treatment, grinding and ultra-precision multi-process. Through the multi-dimensional residual error hierarchical coupling and decoupling algorithm, the module realizes the hierarchical decoupling of independent errors of single process and coupled errors of multiple processes, quantifies the total coupling hazard of errors, and extracts the features of hidden defects. The individual process endowment adaptive threshold quality inspection judgment module is linked with the multi-process residual error coupling and decoupling feature extraction module. Based on the quantified error coupling hazard degree and combined with the individual micro-process parameters of the bearing in the preprocessed data, the module realizes personalized quality inspection threshold correction and differentiated quality judgment for a single bearing through the process endowment difference dynamic threshold correction algorithm. The data storage module is used to receive and store the full-link data from the data acquisition module, the multi-process residual error coupling and decoupling feature extraction module, the individual process endowment adaptive threshold quality inspection judgment module, and the result output module. The result output module is used to receive the final data from the data storage module and the individual process endowment adaptive threshold quality inspection judgment module. Based on the calculation results of the coupling decoupling algorithm and the dynamic threshold correction algorithm, it completes the standardized integration, hierarchical output, working condition traceability display and batch optimization result push of bearing quality inspection data.

2. The bearing production quality inspection system based on production data analysis according to claim 1, characterized in that: The specific implementation steps of the multi-process residual error coupling decoupling feature extraction module are as follows: Step A1: Collection and Standardization of Residual Error Data from Multiple Processes: Receive preprocessed data from the acquisition module and collect four categories of original residual error data: forging density deviation of a single bearing, residual deformation from heat treatment, grinding dimensional deviation, and ultra-precision roughness deviation. Remove abnormal extreme values ​​caused by equipment failures. Use an extreme value normalization method to uniformly transform process error data of different dimensions and ranges into standardized residual errors within the 0~1 range. Eliminate the impact of dimensional differences on coupling analysis; Step A2, Statistical Analysis of Process Error Correlation and Coupling Coefficient Calculation: Based on the multi-process error dataset of bearings in the same batch, calculate the covariance and variance of each pair of process errors, substitute them into the coupling coefficient calculation formula, and obtain the error coupling influence coefficient between any two processes. Combined with bearing model matching coupling correction coefficient This allows for adaptive calibration of the coupling coefficient, precisely matching the forming characteristics of different bearings. Step A3, Layered Decoupling and Coupling Hazard Calculation: Using the coupling decoupling formula, the independent error impact value of a single process and the cross-coupling error impact value of two processes are calculated layer by layer, and finally the total coupling hazard of a single bearing is obtained by summing them up. ; Step A4, Feature Output and Data Transmission: The calculated total coupling hazard degree... The independent error and cross-coupled error characteristic data of each process are packaged and output to the individual process endowment adaptive threshold quality inspection judgment module.

3. The bearing production quality inspection system based on production data analysis according to claim 2, characterized in that: The specific implementation steps for the statistical analysis of process error correlation and the calculation of coupling coefficient in step A2 are as follows: Step A21: Construction of multi-process error pairing dataset: Standardize residual errors based on the four core processes of a single bearing. forging, Heat treatment Grinding, Ultra-precision is the basic unit. It collects all sample data from the current complete production batch, constructs a pairwise paired error dataset for each process, and forms... , , , , , There are six sets of process pairings, covering all cross-process error coupling scenarios; Step A22, Solving for statistical characteristic quantities of process error: For each pair of processes... Calculate the covariance of process errors and the variance of single-process errors separately, specifically by setting up statistical rules: solve for the covariance of the two sets of process errors. , used to characterize the , The degree of correlation between residual errors in two processes is considered; a larger value indicates a stronger coupling and correlation between the errors of the two processes. The variance of single-process errors is calculated simultaneously. , It is used to characterize the degree of fluctuation and dispersion of residual error in a single process in batch production, and to reflect the stability of single-process processing; Step A23, Initial Coupling Correlation Coefficient Calculation: Substitute the covariance and variance data of each process group into the uncalibrated basic coupling coefficient formula to calculate the initial coupling correlation coefficient: ;in The initial coupling coefficient is not adapted to the bearing's process characteristics; Step A24, Bearing Model Adaptive Coefficient Calibration: Based on the bearing types of the current production line, match the process coupling correction coefficient. Complete adaptive calibration; Step A25, Final Coupling Influence Coefficient Solution and Consolidation: The initial coupling coefficients are then... Category Adjustment Factor Substitute into the coupling coefficient formula The final error coupling influence coefficients corresponding to the six pairs of process operations were calculated one by one. , The actual coupling hazard weights after the superposition of errors from different processes were quantified, and adaptive calibration of all process coupling parameters was completed. The solved parameters were then used to calculate the total... The parameters are output to step A3.

4. The bearing production quality inspection system based on production data analysis according to claim 3, characterized in that: The process coupling correction coefficient in step A24 The specific calculation logic is as follows: Step A24.1, Calculation of Discrete Entropy of Single-Process Error: To quantify the degree of process fluctuation disorder in the current batch production of each process, a discrete entropy model of process error is constructed to characterize the disordered fluctuation characteristics of residual error in a single process. The formula for calculating the discrete entropy of single-process error is as follows: , specific For the first The discrete entropy of the error of each process, where the process number is... These correspond to forging, heat treatment, grinding, and ultra-precision processes, respectively. For the first Standardization residual error of each process The probability distribution density across all samples in the current batch; The number of intervals for process error statistics; Step A24.2, Batch Global Process Correlation Calculation: Based on the inherent process sequence logic of the four bearing processes, a batch global process correlation coefficient is constructed to quantify the overall linkage coupling strength of the multi-process errors in the entire batch. The calculation formula is as follows: ;in This is the batch-wide process correlation coefficient, with a value range of [value range missing]. ; Step A24.3, Dynamic Coupling Correction Coefficient Fusion Solution: Combining the discrete entropy of single-process fluctuation characteristics with the correlation degree of global process linkage characteristics, an adaptive fusion algorithm is constructed to dynamically solve the coupling correction coefficient adapted to the current batch production conditions. Core calculation formula: ;in The mean of the discrete entropy of the errors in the four processes satisfies It is used to characterize the overall process fluctuation level of the current batch of bearings; the constant 1.0 is the base correction value to ensure the stability of the parameter base; accomplish The value is dynamically and adaptively adjusted according to the operating conditions, and eventually stabilizes at a certain value. Interval.

5. The bearing production quality inspection system based on production data analysis according to claim 2, characterized in that: The specific implementation steps of step A3 are as follows: Step A31, Core Calculation Parameter Collection and Validity Verification: First, collect the standardized residual errors of the four processes. , , , Independent error influence coefficients for each process The dynamic coupling influence coefficient of the six process pairings ; Step A32, Calculation of Independent Hazardous Component of Single Process Error: First, the independent hazard effect of residual error in a single process is isolated. This component represents the basic hazard of a small residual error in a single process to bearing quality, without cross-process superposition effect. The calculation formula is: ;in, This represents the total hazard component of independent errors in a single process. The preset single-process independent error influence coefficient is used to characterize the basic hazard weight of errors in forging, heat treatment, grinding, and ultra-precision single processes; This refers to the standardized residual error of the corresponding process. Step A33, Calculation of the Hazardous Component of Cross-Process Coupling Errors: Calculate the additional hazard component caused by the superposition of errors across processes. The calculation formula is as follows: In the formula, This refers to the component of errors caused by cross-coupling of multiple processes; The dynamic process coupling influence coefficient, which is adaptively solved in step A2, integrates the entropy value of the current batch process fluctuation and the process correlation characteristics, and can accurately adapt to real-time production conditions. It represents the coupled superposition of residual errors in two processes, characterizing the synergistic superposition effect of minute errors in different processes; Step A34: Calculation of Total Coupling Hazard: Combining the basic hazard components of a single process with the superimposed hazard components of multiple processes, the total coupling hazard is obtained, fully characterizing the overall defect risk of the bearing. Substituting this into the coupling decoupling formula: ;in The total coupling hazard of residual errors in multiple processes of bearing is defined, and its value is greater than or equal to 0. Step A35, Latent Defect Risk Quantification and Calibration: Based on the solution obtained... Complete defect risk stratification and calibration: Single-process error meets standards but Bearings exceeding the adaptive threshold are identified as risk samples with implicit coupling defects.

6. The bearing production quality inspection system based on production data analysis according to claim 1, characterized in that: The specific implementation steps of the individual process endowment adaptive threshold quality inspection judgment module are as follows: Step B1, Individual Process Parameter Collection and Endowment Index Calculation: Collect microscopic process data from the entire production process of a single bearing: raw material crystal phase uniformity test data, machining tool clamping pressure fluctuation data, and machine tool instantaneous machining accuracy data. After standardizing and scoring these three types of data, substitute them into the endowment index calculation formula to obtain the process endowment index specific to each single bearing. It accurately characterizes an individual's tolerance to error. Step B2, Adaptive Quality Inspection Threshold Dynamic Correction: Retrieve Industry Standard Fixed Thresholds The coupling hazard degree transmitted by the multi-process residual error coupling decoupling feature extraction module is combined with the feature extraction module. The endowment index calculated in step B1 The personalized quality inspection threshold for a single bearing is calculated using a dynamic threshold correction formula. ; Step B3, Differentiated Quality Grade Determination: Compare the actual process error and coupling error of a single bearing with the personalized correction threshold, and preset four-level judgment standards: high-quality product, qualified product, warning product, and unqualified product; Step B4, Data Reverse Iteration and Result Output: Output the quality inspection judgment result of a single bearing, and simultaneously store the individual bearing endowment parameters, threshold correction data, and coupling error matching data. Iterate backwards to optimize the process coupling correction coefficient of the multi-process residual error coupling decoupling feature extraction module. This improves the decoupling accuracy of coupling errors in subsequent batches.

7. The bearing production quality inspection system based on production data analysis according to claim 6, characterized in that: The specific implementation steps for collecting individual process parameters and calculating the endowment index in step B1 are as follows: Step B11, Accurate collection and anomaly filtering of multi-dimensional microscopic process parameters: Targeted collection of three types of core original process data that cause individual differences in the endowment of bearings in the same batch: First, the original test data of the crystal phase uniformity of raw materials, which characterizes the structural uniformity and stability of the bearing substrate. Second, the data on the pressure fluctuation of the tooling clamping is used to characterize the deviation in machining stability during the assembly and clamping process; third, the data on the instantaneous machining accuracy of the machine tool is used to characterize the fluctuation in the equipment's operating accuracy at the moment of machining a single bearing. Step B12, Multidimensional Parameter Normalization and Standardized Scoring: Perform extreme value normalization on the original data, uniformly transforming it into... The standardized scores for each interval yielded the following: raw material phase uniformity score. Tooling clamping stability score Machine tool machining accuracy rating ; Step B13, Adaptive Matching of Process Weight Coefficients: Based on the characteristics of the multi-stage forming process of bearings, assign weight coefficients corresponding to the three types of scores. , , The weight normalization constraint is satisfied: In the formula, The raw material crystal phase weighting coefficient is the highest weighting coefficient because the characteristics of the raw material substrate determine the strength of the bearing basic structure. The tooling clamping stiffness weighting coefficient affects the consistency of grinding and ultra-precision machining processes. It serves as a weighting coefficient for machine tool machining accuracy, reflecting the characteristics of instantaneous machining error fluctuations; Step B14, Weighted Solution of Individual Process Endowment Index: Substitute the standardized scoring parameters and adaptive weight parameters into the endowment index calculation formula to accurately solve the individual process endowment index of a single bearing. The formula is as follows: ; Step B15, Endowment Index Validation and Parameter Alignment: Verify the validity of the obtained endowment index. Perform range validation to remove invalid calculation results that exceed the range; simultaneously, include compliant results... The parameters are cached and the total coupling hazard of the single bearing is transmitted in step A4. This forms a one-to-one matching parameter set.

8. The bearing production quality inspection system based on production data analysis according to claim 1, characterized in that: The specific implementation steps for the adaptive quality inspection threshold dynamic correction in step B2 are as follows: Step B21, Core Correction Parameter Collection and Validity Verification: Collect the industry standard fixed qualified thresholds used in this threshold calculation. Fixed threshold correction coefficient Individual bearing process endowment index Total Hazard of Multi-Process Error Coupling ; Step B22, Calculation using the dynamic threshold correction formula: Substitute all compliant parameters into the dynamic threshold correction formula to solve for the personalized adaptive quality inspection threshold for a single bearing. The core formula is as follows: ;in The corrected individual adaptive quality inspection threshold; Set a fixed pass threshold for industry standards; This is the threshold correction coefficient, fixed at 0.3, to control the threshold correction magnitude and avoid excessive deviation from the standard; the formula is... To differentiate the quality of individual bearings, a threshold of 0.5 is used to distinguish between high-tolerance and low-tolerance bearings; through... The coupling defect risk of the multi-process residual error coupling decoupling feature extraction module is linked to achieve bidirectional coupling correction of individual endowment characteristics and implicit coupling risks; Step B23, Implementation of scenario-based adaptive threshold correction logic: Based on step 22 The calculation results, combined with the actual production characteristics of the bearing, yielded four precisely tailored threshold correction conditions: Operating Condition 1: High-Endowment, Low-Risk, High-Quality Bearings: When and At that time, the bearing raw materials were uniform, the processing stability was high, and there were no hidden defects caused by the coupling of errors from multiple processes. After the formula calculation... Expand the quality inspection threshold; Operating Condition 2: High-Endowment, High-Risk Coupled Bearings: When and At that time, the bearing has excellent basic tolerance, but there is a significant hidden danger of multi-process error coupling and superposition. After the formula calculation, the threshold dropped slightly, the judgment standard was appropriately tightened, the hidden coupling defects were accurately identified, and the inherent advantages were avoided from masking the risk of superposition failure. Operating Condition 3: Low-endowment, low-risk general bearings: when and At that time, the individual structural stability and error tolerance of the bearing were relatively weak, but there were no hidden coupling defects. After formula calculation... Appropriately tighten the threshold, strictly control the basic machining accuracy of low-endowment bearings, and avoid the risk of failure in later service life in advance. Operating Condition 4: Low-endowment, high-risk, weak bearings: when and At the same time, bearings simultaneously suffer from individual process defects and multi-process error coupling risks, resulting in a double failure risk. After formula calculation, the threshold is significantly tightened, and strict quality inspection standards are adopted to accurately intercept high-risk defective products. Step B24, Threshold Boundary Constraints and Compliance Calibration: Apply the obtained personalized threshold... Apply boundary constraints to limit the corrected threshold range. ; Step B25, Personalized Threshold Fixing and Parameter Binding: Apply the calibrated single-bearing-specific adaptive quality inspection threshold... With the current bearing , The parameters are bound and stored in a fixed manner to form a unique quality inspection and judgment standard for each bearing, providing a unique and accurate quantitative basis for the next step of differentiated quality level judgment.

Images