Real-time dynamic correction method for screw machining and system thereof

Abstract

The application discloses a real-time dynamic correction method and system for screw machining, which provides a reliable benchmark for correction by designing parameter classification analysis, historical case adaptive optimization of a preset model, and eliminating model deviation by combining error calibration algorithms; multi-dimensional real-time monitoring and scientific data preprocessing ensure that the monitoring data is real and effective, and key feature screening and quantitative calibration can timely warn slight deviation, avoid deviation accumulation, and combine targeted deviation compensation and closed-loop correction to completely solve various processing deviation problems, significantly improve screw geometric precision, and improve precision parameter compliance rate. Real-time weight adaptive strategy can dynamically adjust deviation weight, adapt to real-time processing state, and adapt to different precision grade screw machining requirements; multi-module collaborative linkage and dynamic correction instruction precise execution can timely inhibit processing fluctuation, avoid quality problems caused by equipment operation and cutting state change, and ensure that the processing process is stable and controllable.

Description

Technical Field

[0001] This invention relates to the field of screw machining technology, and in particular to a real-time dynamic correction method and system for screw machining. Background Technology

[0002] As a core component in mechanical transmission and fluid transport, the machining accuracy of screws directly determines the operating efficiency, stability, and service life of equipment, and they are widely used in key equipment such as machine tools, compressors, and injection molding machines. With the rapid development of industrial automation and precision manufacturing, the market demands for screw machining accuracy, quality consistency, and processing efficiency are constantly increasing, especially for high-precision screws, whose geometric parameters, accuracy parameters, and dynamic state fluctuations during the machining process all have a significant impact on the final machining quality.

[0003] Currently, screw machining mostly adopts the traditional mode of preset process parameters and fixed processing. That is, the machining process is set according to the screw design parameters, and cutting operations are performed according to fixed parameters during the machining process. Only after the machining is completed is the finished product inspected, and if deviations are found, rework is carried out. As the complexity of screw structures increases and the precision requirements become more stringent, factors such as dynamic fluctuations such as changes in cutting force, fluctuations in cutting temperature, and unstable spindle speed, as well as deviations in geometric parameters and precision parameters, tend to accumulate, making it difficult for traditional machining modes to meet the demands of high-precision and high-efficiency machining. To address the aforementioned issues, the industry has gradually introduced monitoring and correction technologies, attempting to compensate for processing deviations by collecting monitoring data during the processing. However, existing technologies still have many shortcomings and have failed to form a comprehensive, precise, and dynamic correction system. Therefore, developing a real-time dynamic correction method and system for screw processing that can monitor, accurately analyze, and dynamically compensate in real time has become an urgent need in the current screw processing field. Summary of the Invention

[0004] To achieve the above objectives, one technical solution adopted by the present invention is: a real-time dynamic correction method for screw machining, the method comprising: Input the machining plan for the screw body to be machined, which includes the design parameters and process requirements of the screw body to be machined; By invoking strategies to analyze the design parameters and process requirements of the screw body to be processed, a preset model of the processed body is obtained; The machining process of the screw body to be machined is dynamically monitored by the monitoring module on the sensor tool, and real-time monitoring data is obtained through data preprocessing strategy. By analyzing real-time monitoring data using feature extraction strategies, the real-time state characteristics of the screw body to be processed can be obtained. Real-time state characteristics are used to construct an actual model of the real-time processing body through modeling strategies; The model deviation analysis strategy is used to compare the real-time workpiece actual model with the workpiece preset model to obtain the model comparison deviation and comprehensive deviation factor. When the comprehensive deviation factor is greater than the deviation threshold, the machining plan of the screw body to be processed and the preset model of the machined body are used as the correction targets. The deviation of the model comparison is dynamically corrected through the deviation compensation strategy to obtain the dynamic correction command. The screw body to be machined is dynamically corrected according to the dynamic correction command until the comprehensive deviation factor is less than the deviation threshold, then the dynamic correction of the screw machining is completed.

[0005] Furthermore, the invocation strategy includes: The system calls up a pre-set screw machining parameter database, classifies and analyzes the design parameters of the screw body to be machined, and archives the analyzed parameters according to geometric parameters and precision parameters. Based on the process requirements in the processing plan, the process parameter case library of similar screw processing in the past is called, and the standard process requirements that match the design parameters and material properties of the screw to be processed in the current process with a greater than the preset proportion are selected. The process parameters in the standard process requirements are then extracted as standard parameters. Call the preset screw machining body basic model template, and adjust the parameters of the basic model template according to the analyzed geometric parameters, accuracy parameters and standard parameters to clarify the key feature nodes in the model and the accuracy benchmark of the model; After calling the error calibration algorithm to make preliminary corrections to the geometric and process deviations in the basic model template, the preset model of the processed body is obtained.

[0006] Furthermore, the data preprocessing strategy includes: The raw monitoring data during the machining process is collected synchronously by the monitoring module on the sensor tool, and each raw monitoring data is timestamped. use The criteria, normalization algorithm, and moving average filtering algorithm are used to screen outliers, normalize, and smooth the collected raw monitoring data to obtain effective real-time monitoring data. The preprocessed real-time monitoring data is classified and archived, and associated with the corresponding machining process, tool number and design parameters to form a structured real-time monitoring dataset.

[0007] Furthermore, the feature extraction strategy includes: The real-time monitoring dataset is classified into processing status features and equipment operation features, and basic features of each type of data are extracted. The processing status features include the peak value and fluctuation amplitude of cutting force, the steady-state value and rate of change of cutting temperature, the peak value of vibration frequency and the distribution of the main frequency. The equipment operation features include the stability of spindle speed and the fluctuation of feed rate. The workpiece detection features include the average surface roughness of screw machining and the initial value of contour deviation. At the same time, the timestamps and processing operation nodes corresponding to each feature are associated to form an initial feature set. The preset feature importance evaluation model is invoked, and the correlation coefficient between each feature in the initial feature set and the screw processing quality is calculated in combination with the precision requirements and process characteristics of screw processing. Key features with correlation coefficients greater than the preset threshold are then selected. The key features after screening are quantified, the quantification standards and threshold ranges of each feature are set, and the design parameters and process requirements of the screw to be processed are called to calibrate each key feature to obtain the normal range of the feature and the abnormal warning threshold. The key features after quantitative calibration are integrated to obtain the real-time state feature set of the screw body to be processed.

[0008] Furthermore, the modeling strategy includes: The real-time state feature set of the screw body to be processed is rendered and fused into the preset model of the processing body to obtain the actual model of the real-time workpiece.

[0009] Furthermore, the model bias analysis strategy includes: The model feature alignment algorithm is called to align the key feature nodes, geometric contours, and accuracy benchmarks of the real-time workpiece actual model with the preset model of the machining body. The geometric parameter deviation vector, accuracy parameter deviation vector, and dynamic state deviation vector of the two models are extracted and normalized to obtain the corresponding normalized deviation values. The real-time weights of various deviations are obtained through a real-time weight adaptive strategy, and then weighted with the corresponding normalized deviation values ​​to obtain a comprehensive deviation factor.

[0010] Furthermore, the real-time weight adaptive strategy includes: calculating the window variance of three types of vector data within a preset window: geometric parameter deviation vector, precision parameter deviation vector, and dynamic state deviation vector, and comparing them with the corresponding fluctuation thresholds; Calculate the real-time reliability of the three types of vector data based on the window variance of the three types of vector data; When the window variance of any data point in the three types of vector data exceeds the corresponding fluctuation threshold, the current fluctuation of that data type is abnormal, and the real-time reliability is reduced to the base reliability. ,in, ; The real-time reliability of the three types of vector data is normalized to obtain the real-time weights of the three types of vector data.

[0011] Furthermore, the deviation compensation strategy includes: prioritizing the three types of deviations according to their real-time weights, with the deviation having the higher real-time weight being compensated first. For geometric parameter deviations, a geometric deviation compensation algorithm is used, with the design parameters of the screw body to be processed as the correction target, to calculate the corresponding geometric correction vector; For the deviation of accuracy parameters, the accuracy compensation vector is calculated by using the accuracy parameter deviation compensation algorithm and taking the process requirements of the screw body to be processed as the correction target. For dynamic state deviations, a dynamic compensation algorithm is used to calculate the real-time dynamic correction vector with the pre-set model of the machining body of the screw body to be processed as the correction target; The geometric correction vector, the precision compensation vector, and the dynamic correction vector constitute the dynamic correction command.

[0012] Another technical solution adopted by the present invention is: a real-time dynamic correction system for screw machining. This system is applied to the above-mentioned real-time dynamic correction method for screw machining. The method includes: a screw clamping mechanism 1, a tool spindle 2, a tool magazine assembly 3, a tool changer 4, a sensor tool 5, and a control center 6. The screw clamping mechanism 1 is set at both ends of the screw body to be processed. It is used to clamp, position and fix the screw body to be processed, ensuring that there is no radial runout and axial displacement of the screw body during the processing, providing a stable workpiece reference for screw processing; and driving the screw body to be processed to perform high-speed rotational motion, providing power support for screw cutting; and adjusting the speed and rotation accuracy in real time according to the dynamic correction command issued by the control center. The tool spindle 2 is located above the machining area of ​​the system. The tool is clamped below the tool spindle 2 and the screw body to be machined is turned and milled by the drive mechanism. The tool magazine assembly 3 is located on one side of the tool spindle and is used to store various tools and sensor tools 5 required for machining the screw. It classifies, stores and manages the tools and provides tool position information for the tool changing actuator. The tool changing actuator 4 is located between the tool spindle 2 and the tool magazine assembly 3. It is used to grab the corresponding tool from the tool magazine assembly according to the instructions of the control center and transfer it to the tool spindle 2 to complete the automatic tool changing. The tool changing actuator 4 includes a tool changing arm 401 and a drive assembly 402, wherein the tool changing arm 401 is mounted on the drive end of the drive assembly 402; The tool changer arm 401 is used to grab the tool and sensor tool 5 in the tool magazine assembly and connect to the tool spindle 2; The drive assembly 402 is used to drive the tool changer arm 401 to complete rotation and extension movements; The sensor tool 5 is equipped with a monitoring module, which is used to dynamically collect raw monitoring data during the machining process of the screw body to be processed, and to timestamp each raw monitoring data. The control center 6 is electrically connected to the screw clamping mechanism 1, the tool spindle 2, the tool magazine assembly 3, the tool changer 4, and the sensor tool, and is used to execute all the strategies in the real-time dynamic correction method for screw machining. Furthermore, the control center 6 includes: The pre-processing input module 601 is used to receive the pre-processing plan of the screw body to be processed, and to parse and store the design parameters and process requirements in the pre-processing plan; The calling module 602 is electrically connected to the pre-plan input module. The calling module 602 has a built-in preset screw processing parameter database, a historical database of similar screw processing process parameter cases, and a basic model template of the screw processing body. The calling module 602 is used to execute the calling strategy, classify and analyze the design parameters and process requirements of the screw body to be processed, match process parameters, adapt and adjust the basic model and calibrate errors, and output the preset model of the processing body. The monitoring module 603 includes a monitoring module on the sensor tool. The monitoring module is used to dynamically collect raw monitoring data during the machining process of the screw body to be processed, and add a timestamp mark to each raw monitoring data. Data preprocessing module 604 is used to execute data preprocessing strategies, employing... Criteria, normalization algorithms, and moving average filtering algorithms are used to screen outliers, normalize, and smooth the raw monitoring data. The processed effective real-time monitoring data are classified, archived, and associated with corresponding processing steps, tool numbers, and design parameters to form a structured real-time monitoring dataset. The feature extraction module 605 is electrically connected to the data preprocessing module 604. The feature extraction module 605 is used to execute the feature extraction strategy, classify the structured real-time monitoring dataset, extract basic features, screen key features, quantize and calibrate it, and output the real-time state feature set of the screw body to be processed. The modeling module 606 is electrically connected to the calling module 602 and the feature extraction module 605 respectively. The modeling module 606 is used to execute the modeling strategy, receive the pre-set model of the workpiece output by the calling module and the real-time state feature set output by the feature extraction module, render and fuse the real-time state feature set into the pre-set model of the workpiece, and construct and output the real-time workpiece actual model. The deviation analysis module 607 is electrically connected to the calling module 602 and the modeling module 606 respectively. The deviation analysis module 607 is used to execute the model deviation analysis strategy and the real-time weight adaptive strategy. It calls the model feature alignment algorithm to realize the alignment of the real-time workpiece actual model with the preset model of the machining body, extracts various deviation vectors and performs normalization processing, calculates the real-time weight of various deviations through the real-time weight adaptive strategy, and then obtains the model comparison deviation and comprehensive deviation factor through weighted calculation. At the same time, it presets the deviation threshold and completes the comparison between the comprehensive deviation factor and the deviation threshold. The compensation module 608 is electrically connected to the deviation analysis module 607. The compensation module 608 is used to execute a deviation compensation strategy. According to the real-time weight output by the deviation analysis module, the three types of deviations are prioritized and sorted. The geometric correction vector, the accuracy compensation vector, and the real-time dynamic correction vector are calculated by the corresponding compensation algorithm. The three types of vectors are integrated to form a dynamic correction instruction. The execution module 609 is electrically connected to the compensation module 608 and the monitoring module 603 respectively. The execution module 609 is used to receive the dynamic correction command output by the compensation module, control the processing equipment to perform dynamic processing correction on the screw body to be processed; and receive the correction process monitoring data fed back by the monitoring and preprocessing module in real time, and cooperate with the deviation analysis module to cyclically calculate the comprehensive deviation factor until the comprehensive deviation factor is less than the deviation threshold, and then stop the correction operation.

[0013] Compared with existing technologies, the present invention has the following advantages: 1. By classifying and analyzing design parameters and adapting historical cases to optimize the preset model, and combining error calibration algorithms to eliminate model deviations, a reliable benchmark is provided for correction; multi-dimensional real-time monitoring and scientific data preprocessing ensure that the monitoring data is true and effective; key feature screening and quantitative calibration can provide timely warnings of slight deviations and avoid the accumulation of deviations; combined with targeted deviation compensation and closed-loop correction, various processing deviation problems are completely solved, and the screw geometric accuracy and accuracy parameter compliance rate are significantly improved.

[0014] 2. The real-time weight adaptive strategy can dynamically adjust the deviation weight to match the real-time processing status and adapt to the processing requirements of screws with different precision levels; the multi-module collaborative linkage and dynamic correction commands are executed accurately, which can suppress processing fluctuations in a timely manner, avoid quality problems caused by changes in equipment operation and cutting conditions, and ensure that the processing process is stable and controllable. Attached Figure Description

[0015] Figure 1 This is a flowchart illustrating the real-time dynamic correction method for screw machining according to the present invention.

[0016] Figure 2 This is a schematic diagram of the module connections in the control center.

[0017] Figure 3This is a schematic diagram of a real-time dynamic correction system used for screw machining. Figure 4 This is a schematic diagram of the screw clamping mechanism; Figure 5 This is a schematic diagram showing the position and structure of the tool changing actuator and the tool spindle.

[0018] Figure 6 This is a schematic diagram of the tool changing actuator.

[0019] The components include: 1. Screw clamping mechanism; 2. Tool spindle; 3. Tool magazine assembly; 4. Tool changer; 401. Tool changer arm; 402. Drive assembly; 5. Sensor tool; 6. Control center; 601. Pre-program input module; 602. Calling module; 603. Monitoring module; 604. Data preprocessing module; 605. Feature extraction module; 606. Modeling module; 607. Deviation analysis module; 608. Compensation module; and 609. Execution module. Detailed Implementation

[0020] The technical solutions of the real-time dynamic correction method and system for screw machining provided by the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0021] Example 1 like Figure 1 As shown, a real-time dynamic correction method for screw machining is described. The method includes: inputting a machining plan for the screw body to be machined, wherein the machining plan includes the design parameters and process requirements of the screw body to be machined.

[0022] Specifically, the design parameters cover core geometric and precision-related parameters such as screw nominal diameter, pitch, thread angle, groove depth, and effective length. The process requirements include machining constraints such as cutting speed, feed rate, depth of cut, tool selection, and machining accuracy level. Their function is to provide a benchmark for the entire screw machining and dynamic correction process, ensuring that the machining process always revolves around the preset target and avoiding machining deviations exceeding the allowable range.

[0023] The method also includes: analyzing the design parameters and process requirements of the screw body to be processed by calling a strategy to obtain a preset model of the processed body.

[0024] Furthermore, the invocation strategy includes: invoking a preset screw machining parameter database, classifying and parsing the design parameters of the screw body to be machined, and classifying and archiving the parsed parameters according to geometric parameters and precision parameters.

[0025] Specifically, the process first calls a pre-set screw machining parameter database to determine if the database contains parameter templates of the same type for the screw body to be machined. If not, a parameter matching warning is issued, prompting staff to supplement the relevant parameter templates. If the database contains the templates, the design parameters of the screw to be machined are analyzed one by one, and the attributes of each parameter are determined. The criteria for determining geometric parameters are those describing the screw's shape and structural dimensions, while the criteria for determining precision parameters are those describing the screw's machining precision and geometric tolerances. After analysis, the classification results are checked for overlap or omissions. If these exist, the classification is re-analyzed. If the classification is clear and without omissions, the two types of parameters are archived separately to form a geometric parameter set. and precision parameter set ,in Representing the Geometric parameters, , Representing the Precision parameters, , , These represent the number of geometric parameters and the number of precision parameters, respectively.

[0026] Based on the process requirements in the processing plan, the process parameter case library of similar screw processing in the past is called up, and the standard process requirements that match the design parameters and material properties of the screw to be processed in the current process with a greater than preset ratio are selected. The process parameters in the standard process requirements are then extracted as standard parameters.

[0027] Specifically, first, extract the core process requirements from the processing plan, including the machining accuracy level and cutting method, as well as the material properties of the screw to be processed, such as hardness. and resilience First, it calls up a historical database of similar screw machining process parameters; second, it calculates the matching degree between each historical case in the database and the current screw to be machined. The matching degree is calculated using the following formula: , in, This represents the matching degree between historical cases and the current screw to be processed, with a value range of [0,1]. The closer the match is to 1, the higher the degree of matching. , , These are weighting coefficients, corresponding to the matching weights of geometric parameters, accuracy parameters, and material properties, respectively. ; , The first in historical cases geometric parameters, the first The value of the precision parameter; The maximum value of the geometric parameters for screws of the same type. The maximum value of the accuracy parameters for screws of the same type is used for normalization to ensure that the calculation dimension of the matching degree of various parameters is consistent. , These are the hardness and toughness values ​​of the screw material in historical cases. This represents the maximum value of characteristic parameters for similar screw materials, used for normalization.

[0028] After the calculation is complete, set the preset matching ratio. To judge each historical case Is it greater than If a certain historical case If the case is deemed a suitable case, then after filtering out all suitable cases, the process parameters of each suitable case are extracted, and the average value of each process parameter is calculated as the standard parameter for the screw to be processed. ,in Representing the Standard process parameters, , The number of standard process parameters; if all historical cases Based on the process requirements of the processing plan and combined with the screw processing parameter database, basic process parameters are generated as standard parameters and marked as cases without historical adaptation, so as to facilitate key monitoring in subsequent processing.

[0029] The preset screw machining body basic model template is called, and the parameters of the basic model template are adjusted according to the analyzed geometric parameters, accuracy parameters and standard parameters to clarify the key feature nodes in the model and the accuracy benchmark of the model.

[0030] Specifically, first, a preset basic model template for the screw machining body is called, and it is determined whether the template matches the type of the screw to be machined. If they do not match, the basic model template of the corresponding type is called; if they match, the geometric parameter set is... Precision parameter set Standard parameter set Input the basic model template one by one, and adjust the corresponding parameters of the template accordingly. After adjustment, determine whether the key feature nodes of the model, such as the vertices of the thread profile, the bottom nodes of the thread groove, and the transition nodes of the steps, are clear and complete, and whether they are consistent with the design parameters. At the same time, clarify the accuracy benchmark of the model. That is, the set of precision parameters Core precision indicators, such as pitch tolerance Surface roughness If key feature nodes are missing, unclear, or the accuracy benchmark is unclear, the parameters are readjusted until the model meets the requirements of clear features, clear benchmark, and matching parameters, thus obtaining a preliminary adapted model.

[0031] After calling the error calibration algorithm to make preliminary corrections to the geometric and process deviations in the basic model template, the preset model of the processed body is obtained.

[0032] Specifically, the error calibration algorithm is first invoked to calculate the geometric deviations in the preliminary adaptation model. and process deviation The geometric deviation is calculated using the least squares method, and the formula is as follows: , in, For geometric deviation, the smaller the value, the higher the geometric accuracy; To initially adapt the model in the first The numerical values ​​of the geometric parameters; The first parameter in the design parameters of the screw to be processed Standard values ​​of the geometric parameters; The number of geometric parameters.

[0033] Process deviation The calculation uses the deviation accumulation algorithm, and the formula is: , in, For process deviation, the smaller the value, the better the process adaptability; For the first The weights of the standard process parameters are set according to the degree of influence of the process parameters on the processing quality, and the sum of the weights is 1. To initially adapt the model in the first The values ​​of the process parameters; For the first in the standard parameter set The values ​​of the process parameters; This refers to the number of standard process parameters.

[0034] Calculated and Then, set the geometric deviation threshold. and process deviation threshold (Based on the precision requirements set in the processing plan), perform logical judgment: (1) If and If the initial adaptation model deviation is deemed acceptable, no further correction is needed, and the model can be directly used as the preset model for the processing body.

[0035] (2) If and Then only the geometric parameters are corrected, based on the geometric deviation. The size is adjusted proportionally to the geometric parameters in the model, and the corrected formula is: ,in For the revised version The numerical values ​​of the geometric parameters, This is the geometric correction factor. , The larger, The closer it is to 1, the more it needs to be corrected and recalculated. until .

[0036] (3) If and In this case, only the process parameters are corrected, based on the process deviation. Adjust the size of the process parameters in the model, and correct the formula as follows: ,in For the revised version The values ​​of the process parameters, Here, is the process correction factor, , The larger, The closer it is to 1, the more it needs to be corrected and recalculated. until .

[0037] (4) If and Then, both geometric and process parameters are corrected simultaneously, using the aforementioned geometric and process correction formulas respectively. Correction priorities are assigned according to the magnitude of the deviation; that is, the larger the deviation, the higher the correction priority. After correction, the calculation is repeated. and This continues until both deviations meet the threshold requirements. After the correction is completed, the key feature nodes and accuracy benchmarks of the model are checked again to see if they remain clear and consistent. If they meet the requirements, the model is output as the preset model of the machining body. If they do not meet the requirements, the parameters are adapted and the error is calibrated again to ensure the accuracy and reliability of the preset model of the machining body, laying the foundation for the subsequent real-time dynamic correction process.

[0038] In this embodiment, by classifying and analyzing design parameters, screening suitable process parameters, and optimizing the selection of standard parameters based on historical cases, the adaptability and rationality of model parameters are improved. This solves the problems of poor reusability and low adaptability of similar screw processing parameters and reduces the blindness in setting process parameters. The error calibration process calculates geometric and process deviations and makes targeted corrections to avoid dynamic correction inaccuracies caused by preset model deviations.

[0039] The method also includes: dynamically monitoring the machining process of the screw body to be machined through the monitoring module on the sensor tool, and obtaining real-time monitoring data through a data preprocessing strategy.

[0040] Furthermore, the data preprocessing strategy includes: synchronously collecting raw monitoring data during the machining process through the monitoring module on the sensor tool, and simultaneously timestamping each raw monitoring data.

[0041] Specifically, the monitoring module on the sensor tool is first activated, which simultaneously triggers multiple types of sensors, such as cutting force sensors, temperature sensors, vibration sensors, and displacement sensors.

[0042] Collect raw monitoring data set ,in , Representing the Similar to raw monitoring data, This represents the total number of sensor types. Secondly, for each type of raw monitoring data Add timestamps to generate a raw dataset with timestamps. ,in For the first The collection timestamp corresponding to the data type; Finally, determine whether the collected raw data contains invalid cases such as all zeros or no change. If a certain type of data... If the value remains 0 or unchanged, an abnormal data acquisition warning will be issued, prompting the user to check whether the corresponding sensor is faulty. At the same time, invalid data of this type will be discarded, and only valid data will be retained for the next step of preprocessing.

[0043] use Criteria, normalization algorithms, and moving average filtering algorithms are used to screen outliers, normalize, and smooth the collected raw monitoring data to obtain effective real-time monitoring data.

[0044] Specifically, for each type of raw monitoring data with timestamps First, invalid data from the initial screening are removed, and then the mean is calculated. and standard deviation Extreme outliers that exceed reasonable ranges are filtered out based on statistical characteristics to ensure that the retained data conforms to the normal fluctuation patterns of the processing.

[0045] After screening, calculate the percentage of outliers. ,in, To determine the number of outliers, perform a double logical check: if ,in, If the preset percentage is not met, the sensor's data acquisition stability is deemed insufficient, a sensor calibration prompt is issued, and adjacent valid data interpolation is used to fill in the missing outlier positions to ensure data continuity; if If the collected data is stable, the data after removing outliers will proceed to the next step of normalization processing.

[0046] After outlier screening, the valid data exhibits significant differences in the units and ranges of various indicators. Directly using these data for feature extraction and bias analysis would lead to weight imbalance. Therefore, all data must be uniformly mapped to the [0,1] interval to eliminate the differences in units between data points and ensure the fairness and accuracy of subsequent analysis. The calculation formula is as follows: , in, For the first After the data of class 1 is normalized, the first class 2 is... The values ​​of each valid collection point range from [0,1]. The closer the value is to 1, the closer the monitoring indicator is to its maximum value. For the first The minimum value of valid raw data after outlier screening reflects the lowest monitoring value of this type of indicator; For the first The maximum value of the valid raw data after outlier screening reflects the highest monitoring value of that type of indicator. For the first The valid raw data after outlier screening is the input data for normalization.

[0047] Even after normalization, the data stream still contains high-frequency noise, which can obscure the true trend of the processing status. Therefore, a moving average filtering algorithm is used to smooth the data through a sliding window, filtering out high-frequency noise and restoring the true changing patterns of the processing status, thus providing a stable data foundation for subsequent feature extraction. The calculation formula is as follows: , in, For the first After smoothing, the first class of data... The final valid values ​​from each collection point reflect the true trend of the processing status; This is the size of the filtering window; The radius is the window radius, rounded down. For the first In normalized data, the first class Centered on the nth point, the nth point within the window range The data from each sampling point is used to calculate the smoothing value of the current point.

[0048] The preprocessed real-time monitoring data is classified and archived, and associated with the corresponding machining process, tool number and design parameters to form a structured real-time monitoring dataset.

[0049] Specifically, the core information of the screw to be processed is extracted from the processing plan, including: the set of geometric parameters G and the set of precision parameters. For example, rough machining, semi-finishing, and finishing are denoted as... At the same time, the time interval for each process and the tool number information currently used in the tool magazine assembly are clearly defined and recorded as follows: Determine if the extracted information is complete. If there are problems such as missing design parameters or unclear process information, suspend the preprocessing process and send feedback to the control center to supplement and improve the processing plan.

[0050] Effective real-time monitoring data after smoothing Perform classification mapping: based on the timestamps corresponding to the data. Match it to the processing procedure it is in ,like During the finishing stage, it is mapped to... At the same time, associate it with the tool number used in the current process. and screw design parameters , Determine if the mapping is accurate. If there is a mismatch between the timestamp and the process time interval, or an incorrect tool number association, remap the data to ensure that the data corresponds to the scenario.

[0051] Building a structured real-time monitoring dataset Its data structure is defined as follows: .

[0052] In this embodiment, the accuracy and reliability of dynamic correction are further improved through real-time monitoring and data preprocessing of the manufacturing process. Simultaneous acquisition by multiple sensors combined with timestamp marking enables full-process traceability of monitoring data, initial rejection of invalid data, and timely warning of sensor malfunctions, ensuring the effectiveness of data acquisition. The criteria, normalization, and moving average filtering algorithms effectively screen outliers, eliminate dimensional differences, filter high-frequency noise, restore the true trend of the processing status, and avoid data distortion interfering with subsequent analysis. Structured archiving and multi-information association ensure that the data corresponds to the processing scenario, providing stable and standardized high-quality data support for feature extraction and deviation analysis, and helping to improve the consistency of screw processing accuracy and quality.

[0053] The method also includes: analyzing real-time monitoring data through feature extraction strategies to obtain the real-time state characteristics of the screw body to be processed.

[0054] Furthermore, the feature extraction strategy includes: classifying the real-time monitoring dataset into processing status features and equipment operation features, and extracting basic features for each type of data. The processing status features include the peak value and fluctuation amplitude of cutting force, the steady-state value and rate of change of cutting temperature, the peak value of vibration frequency and the distribution of the dominant frequency. The equipment operation features include the stability of spindle speed and the fluctuation of feed rate. The workpiece detection features include the average surface roughness of screw machining and the initial value of contour deviation. At the same time, the timestamps and processing process nodes corresponding to each feature are associated to form an initial feature set.

[0055] Specifically, input a structured real-time monitoring dataset: , First, input validation is performed to determine if the dataset is complete, i.e., there are no missing processes, no blank monitoring data, and no incorrect timestamps. If there is missing or incorrect data, the process is traced back to the data preprocessing stage to supplement and improve it, ensuring the validity of the input data. If the data is complete, the process proceeds to the classification stage.

[0056] Secondly, based on data attributes and their correlation with processing quality, classification rules are established for... The various monitoring data are categorized, and the logical judgment criteria are as follows: If the monitoring data directly reflects the mechanical, thermal, and vibrational states during the cutting process, it is classified as machining state characteristics and denoted as a set. ; If the monitoring data reflects the operating parameters of the machining equipment spindle and feed mechanism, such as spindle speed and feed rate, then it is classified as equipment operating characteristics and denoted as a set. ; If the monitoring data directly reflects the surface quality, surface roughness, and contour deviation of the screw after machining, it is classified as a workpiece inspection feature and denoted as a set. .

[0057] Finally, for the three types of feature sets, the corresponding basic statistical features and feature indicators are extracted respectively. The specific extraction logic and judgment are as follows: Processing status characteristics Extracting peak cutting force Cutting force fluctuation range Steady-state value of cutting temperature Cutting temperature change rate Peak vibration frequency Vibration dominant frequency distribution The fluctuation amplitude and rate of change are obtained through statistical calculations, and the formula is: Cutting force fluctuation amplitude: ,in This represents the maximum cutting force within a specific machining operation. This represents the minimum cutting force during this process; the rate of change of cutting temperature is: ,in for Cutting temperature at any given time for Cutting temperature at any given time , These are the timestamps for the corresponding moments. Equipment operating characteristics. Extracting spindle speed stability Feed rate fluctuation The spindle speed stability is quantified using the standard deviation, and the formula is as follows: ,in For a certain process, the first The spindle speed at each acquisition point This represents the average spindle speed during this process. This represents the total number of rotational speed data collection points within this process; the formula for feed rate fluctuation is the same as the cutting force fluctuation amplitude, i.e. , , These represent the maximum and minimum feed rates within a specific process.

[0058] Workpiece inspection features Extract the average surface roughness. Initial value of profile deviation The formula for the average surface roughness is: ,in For the first Surface roughness values ​​at each sampling point Total number of surface roughness sampling points; initial value of profile deviation. This represents the average initial deviation between the actual profile and the designed profile of the screw during a certain process.

[0059] Finally, each of the extracted basic features is associated with its corresponding timestamp. and processing steps This ensures that each feature can be traced back to a specific processing moment and stage, forming an initial feature set. After association, a verification is performed to determine if there are any errors in the association between features and timestamps or process nodes. If such errors exist, the association is re-associated to ensure accuracy, and finally, a complete initial feature set is output.

[0060] The feature extraction strategy also includes: calling a preset feature importance evaluation model, combining the precision requirements and process characteristics of screw processing, calculating the correlation coefficient between each feature in the initial feature set and the screw processing quality, and screening out key features with a correlation coefficient greater than a preset threshold.

[0061] Specifically, the pre-defined feature importance assessment model is first invoked, which is built based on the random forest algorithm. At the same time, the core quality indicators of screw machining are extracted from the machining plan. This serves as a reference benchmark for correlation analysis; it helps determine whether the evaluation model is being used correctly and its quality indicators. Check for completeness; if the model call fails, restart the call; if quality indicators are missing, supplement and improve them.

[0062] Secondly, the Pearson correlation coefficient is used to quantify the initial feature set. Each feature Processing quality indicators The degree of linear correlation between the feature and the processing quality is determined by the absolute value of the correlation coefficient. The larger the absolute value of the correlation coefficient, the stronger the correlation between the feature and the processing quality. The formula is: , in, For the first Initial features Processing quality indicators The Pearson correlation coefficient has a range of [-1, 1]; For the first Initial features The One collected value; For the first Initial features The arithmetic mean of all collected values; For processing quality indicators In the The values ​​corresponding to each data collection time; For processing quality indicators The arithmetic mean of all collected values; This represents the total number of synchronous data collection points for features and quality indicators.

[0063] Finally, based on the precision requirements of screw machining, a preset threshold for the correlation coefficient is set. The higher the threshold, the stronger the correlation between the selected features and the processing quality; each initial feature is analyzed one by one. Make a judgment.

[0064] like If the feature is determined to be a key feature strongly correlated with processing quality, it will be included in the candidate key feature set. ; like If the feature is redundant, it will be discarded.

[0065] The key features selected are quantified, and the quantification standards and threshold ranges of each feature are set. The design parameters and process requirements of the screw to be processed are called to calibrate each key feature and obtain the normal range and abnormal warning threshold of the feature.

[0066] Specifically, firstly, regarding the set of key features Each feature in First, determine whether the data type is continuous or discrete, and then use the corresponding quantization method to uniformly map all features to the [0,10] interval.

[0067] Continuous characteristics such as peak cutting force, cutting temperature, and surface roughness are quantized using the gradient quantization method, with the following formula: , in, For the first The quantized values ​​of the key features range from [0,10]. For the first The original collected values ​​of the key features; For the first The minimum theoretical value of each key feature; For the first The maximum theoretical value of each key feature.

[0068] Discrete features such as spindle speed stability level and whether vibration frequency is abnormal are classified and assigned values ​​using a classification method. Fixed quantization values ​​are assigned according to the feature state. The quantization value setting needs to be combined with process requirements to ensure the distinguishability of different states.

[0069] Secondly, the design parameter set of the screw to be processed is called. Set of process requirements Combined with the set of key quantitative features Each quantitative feature is calibrated, its normal range and abnormal warning threshold are defined, and the logical judgment is as follows: Extract the standard values ​​corresponding to each key feature from the design parameters and process requirements, and convert the standard values ​​into corresponding quantitative values. .

[0070] Setting the normal range for features: based on quantization standard values Based on the screw machining accuracy requirements, a normal fluctuation coefficient is set. The formula for the normal range is: .

[0071] Setting anomaly warning thresholds: Anomaly warning thresholds are divided into primary warning thresholds and secondary warning thresholds, and the formula is as follows: Primary warning threshold is... , The level 2 warning threshold is , in, This is the normal fluctuation coefficient. These are the quantitative standard values ​​for key characteristics; Level 1 warnings correspond to minor deviations and require close monitoring; Level 2 warnings correspond to severe deviations and require immediate triggering of the correction process.

[0072] The key features after quantitative calibration are integrated to obtain the real-time state feature set of the screw body to be processed.

[0073] Specifically, key features will be quantified. The corresponding normal range Anomaly warning threshold, timestamp Processing steps Tool number and screw design parameters Process requirements By performing correlation and integration, a structured data format for real-time status features is constructed.

[0074] Determine if the integrated feature set has issues such as missing information or incorrect associations; if so, go back to the corresponding step to supplement and improve it; if not, proceed to the status judgment step.

[0075] Based on the comparison of the quantified values ​​of each key feature with the normal range and warning threshold, the current screw processing status is initially determined, and the status determination results are included in the feature set to facilitate the rapid location of the abnormal type during subsequent deviation analysis.

[0076] like If so, then the characteristic state is normal; like If so, the characteristic state is slightly abnormal and should be marked as requiring monitoring; like If the condition is abnormal, the feature is considered severely abnormal and is marked as requiring immediate correction.

[0077] Output real-time status feature set: After completing all integration, verification, and status judgment, output the real-time status feature set of the screw body to be processed. This set can be directly input into the modeling strategy to build a real-time model of the processing body.

[0078] In this embodiment, the feature extraction strategy further enhances the targeting and scientific rigor of dynamic correction. By classifying features into three categories—processing status, equipment operation, and workpiece inspection—the core indicators of each feature are clearly defined and associated with the processing scenario, achieving full-process traceability of features and avoiding analytical biases caused by feature confusion. Key features are screened based on a random forest model and Pearson correlation coefficient, effectively eliminating redundant information and focusing on core indicators strongly correlated with processing quality, reducing subsequent analysis complexity and improving efficiency. Quantitative calibration and early warning threshold setting can quickly determine the level of feature anomalies, providing early warnings of minor deviations and timely responses to severe deviations, further ensuring the effectiveness and timeliness of dynamic correction and contributing to stable screw processing quality.

[0079] The method also includes: using real-time state features to construct an actual model of the real-time processing body through a modeling strategy.

[0080] Furthermore, the modeling strategy includes: rendering and fusing the real-time state feature set of the screw body to be processed into the preset model of the processing body to obtain the real-time workpiece actual model.

[0081] Specifically, firstly based on the real-time state feature set Processing steps in timestamp First, establish a correspondence between real-time features and preset model feature nodes. After mapping, determine if the correspondence is accurate. If there is a mismatch between features and model nodes, readjust the mapping rules to ensure that each real-time feature corresponds to a relevant node in the preset model.

[0082] Machining characteristics such as cutting force and cutting temperature are mapped to feature nodes of the cutting area, such as the thread cutting surface and the bottom of the thread groove, in the preset model. Equipment operation characteristics such as spindle speed stability and feed rate fluctuation are mapped to the pitch machining node associated with spindle speed, the cutting depth node associated with feed rate, and the process parameter association node in the preset model. Workpiece inspection characteristics such as surface roughness and contour deviation are mapped to the surface quality inspection node, contour accuracy inspection node, and accuracy inspection node in the preset model.

[0083] Real-time feature rendering fusion calculation: A weighted fusion algorithm is used to integrate the quantized values ​​of real-time state features into the corresponding nodes of the preset model to achieve dynamic rendering fusion. The formula is as follows: , in, For the actual model of the real-time machining body in spatial coordinates Model parameter values ​​at the location; Pre-set the model of the workpiece in spatial coordinates The basic parameter values ​​at that location; These are the base weights of the preset model; For the first The quantized values ​​of several real-time key features are taken from... In ; For the first The fusion weights of each real-time key feature are set based on the correlation between the feature and the model node, and satisfy the following conditions: ,in, The number of key features in real time; the higher the correlation, the greater the weight. The coordinates of the screw machining body cover all feature nodes of the screw.

[0084] During fusion computing, it is performed according to the processing step nodes. The calculation is performed in segments, with the feature weights within the same process remaining consistent, while the weights for different processes can be adjusted according to process requirements.

[0085] In this embodiment, the modeling strategy constructs a real-time model of the machining body, providing a real-time reference for dynamic correction and further enhancing the scientific rigor and relevance of the correction. By establishing a mapping between real-time state features and the feature nodes of the pre-set model of the machining body, and combining machining processes and timestamps to achieve a one-to-one correspondence between features and model nodes, model distortion caused by mapping misalignment is avoided, ensuring that real-time features can be integrated into the pre-set model. A weighted fusion algorithm is adopted to scientifically allocate the fusion weights of the pre-set model and various real-time key features, adjusting the weight ratio based on feature correlation, and performing fusion calculations in segments according to processes, taking into account both the stability of the pre-set benchmark and the dynamism of real-time features. The dynamic rendering and fusion of real-time features can restore the real-time state of screw machining, enabling the real-time model to realistically reflect the mechanics, equipment operation, and workpiece quality changes during the machining process.

[0086] The method also includes: comparing the real-time workpiece actual model with the workpiece preset model through a model deviation analysis strategy to obtain the model comparison deviation and comprehensive deviation factor.

[0087] Furthermore, the model bias analysis strategy includes: The model feature alignment algorithm is invoked to align the key feature nodes, geometric contours, and accuracy benchmarks of the real-time workpiece actual model with the preset model of the machining body. The geometric parameter deviation vector, accuracy parameter deviation vector, and dynamic state deviation vector of the two models are extracted and normalized to obtain the corresponding normalized deviation values.

[0088] Specifically, the model feature alignment algorithm is first invoked. This algorithm can be based on the Iterative Closest Point (ICP) optimization algorithm to align the real-time workpiece model. With the pre-set model of the processing body Perform feature alignment.

[0089] The key feature nodes (such as thread apex, thread groove bottom, step end face), geometric contours (such as pitch, thread angle, outer diameter), and accuracy benchmarks (such as surface roughness benchmark and contour accuracy benchmark) of the two models are aligned to ensure that the benchmarks for comparison are consistent.

[0090] Calculate the overlap of the aligned models The formula is: ,in This represents the number of overlapping feature nodes between the two models after alignment. Set the total number of feature nodes for the preset model; set the overlap threshold. ,like If the alignment is correct, then the alignment is acceptable; if If the alignment algorithm fails, it is called again, and alignment parameters such as the number of iterations and the convergence threshold are adjusted. The alignment is repeated until the overlap meets the standard. Subsequently, based on the two aligned models, the geometric parameter deviation vector, the accuracy parameter deviation vector, and the dynamic state deviation vector are extracted respectively.

[0091] Geometric parameter deviation vector Extract the deviations between the two models in key geometric parameters, covering core geometric indicators such as pitch, thread angle, outer diameter, and thread groove depth. The vector expression is as follows: ,in For the first The deviation values ​​of each geometric parameter are calculated as follows: ,in, For the real-time model One geometric parameter value, For the preset model number Each geometric parameter value.

[0092] Precision parameter deviation vector Extract the deviation between the two models in terms of accuracy parameters, covering core accuracy indicators such as surface roughness, contour deviation, and positional accuracy. The vector expression is as follows: ,in For the first The deviation value of each precision parameter is calculated as follows: ,in, For the real-time model A precision parameter value, For the preset model number Each precision parameter value.

[0093] Dynamic state deviation vector Extract the deviation between the two models in terms of dynamic state parameters, covering dynamic indicators such as cutting force fluctuation deviation, cutting temperature deviation, and spindle speed stability deviation. The vector expression is as follows: ,in For the first The deviation value of each dynamic state parameter is calculated as follows: ,in, For the real-time model A dynamic state parameter value, For the preset model number One dynamic state parameter value; Deviation vector normalization: Since the three types of deviation vectors have different dimensions, they cannot be directly weighted. Therefore, it is necessary to normalize the three types of deviation vectors, mapping all deviation values ​​to the [0,1] interval to obtain the corresponding normalized deviation values. The formula is as follows: Geometric parameter normalized deviation values: ; Normalized deviation value of accuracy parameter: ; Dynamic state normalized deviation value: ; in, , , These are the normalized deviation values ​​for geometric parameters, accuracy parameters, and dynamic state parameters, respectively, with a value range of [0,1]. The larger the value, the more severe the corresponding deviation. , , These are the absolute values ​​of the deviation values ​​of each parameter in the three types of deviation vectors; , , These are the maximum permissible deviation values ​​for the three types of deviation parameters.

[0094] Determine if all normalized deviation values ​​are within the range [0,1]. If any values ​​are outside this range, it indicates that the maximum allowable deviation value for the corresponding parameter is not set appropriately. Readjust and re-normalize until all normalized deviation values ​​meet the requirements, thus obtaining the normalized deviation vector. , , .

[0095] The method also includes: obtaining the real-time weights of various deviations through a real-time weight adaptive strategy, and performing weighted calculations with the corresponding normalized deviation values ​​to obtain a comprehensive deviation factor.

[0096] Based on the real-time weight adaptive strategy, the real-time geometric parameter deviation weights of the three types of deviations are obtained. Precision parameter deviation weight Dynamic state deviation weight And satisfy The comprehensive deviation factor is obtained by weighting the real-time weights with the corresponding normalized deviation values, using the following formula: , in, This is the overall deviation factor, with a value range of [0,1]. The closer it is to 1, the more serious the overall processing deviation; the closer it is to 0, the closer the processing state is to the preset standard. , , The real-time weights for the three types of deviations are output by a real-time weight adaptive strategy, reflecting the degree of influence of each type of deviation on the overall processing quality. The average value of the normalized deviation of the geometric parameters. The number of geometric parameters; This is the average value of the normalized deviation of the accuracy parameter. The number of precision parameters; This is the average value of the normalized deviation of the dynamic state. This represents the number of dynamic state parameters.

[0097] Furthermore, the real-time weight adaptive strategy includes: calculating the window variance of three types of vector data—geometric parameter deviation vector, precision parameter deviation vector, and dynamic state deviation vector—within a preset window, and comparing them with the corresponding fluctuation thresholds.

[0098] Specifically, set a preset time window. Collect the geometric parameter deviation vector within the preset window. Precision parameter deviation vector Dynamic state deviation vector Three types of vector data, The time of data collection within the window. , Given the number of data collections within the window, calculate the window variance for each of the three types of vector data. A larger variance indicates more drastic fluctuations in that type of bias, and lower data reliability. The formula is: Geometric parameter deviation vector window variance: , Precision parameter deviation vector window variance: , Dynamic state deviation vector window variance: , in , , These are the window variances of the three types of bias vectors, with values ​​ranging from [value range missing]. The larger the variance, the more drastic the fluctuation. , , They are respectively The values ​​of the three types of deviation vectors at time; , , These are the average values ​​of the three types of deviation vectors within the preset window; This represents the number of times deviation data is collected within a preset window.

[0099] Set fluctuation thresholds corresponding to the three types of deviation vectors. , , The core function of the fluctuation threshold is to define the normal or abnormal boundary of the fluctuation of deviation data. The threshold setting needs to be combined with the screw machining accuracy level. The threshold setting value is smaller for high-precision screws and can be appropriately relaxed for ordinary precision screws.

[0100] The window variance of each of the three types of deviation vectors is compared with the corresponding fluctuation threshold, i.e., each is judged separately. and , and , and The size relationship.

[0101] If the window variance of a certain type of deviation vector is less than or equal to the corresponding fluctuation threshold, such as If the fluctuation of this type of deviation data is normal, the data reliability is at a normal level, and there is no need to adjust its basic reliability; if the window variance of a certain type of deviation vector is greater than the corresponding fluctuation threshold, such as If the data is found to be fluctuating abnormally, its reliability is reduced, and its real-time reliability needs to be adjusted according to the rules.

[0102] The real-time weight adaptive strategy also includes: calculating the real-time reliability of the three types of vector data based on the window variance of the three types of vector data.

[0103] Specifically, the basic confidence levels of the three types of deviation vectors are first set. By combining the comparison results of window variance and fluctuation threshold, the real-time reliability of the three types of data is calculated.

[0104] When the window variance of a certain type of deviation vector is less than or equal to the corresponding fluctuation threshold, the fluctuation of this type of data is normal, and the real-time reliability is equal to its basic reliability, as shown in the formula: , in, For the first Real-time reliability of class bias, For the first The basic confidence level of class bias These correspond to geometric, accuracy, and dynamic state deviations, respectively.

[0105] The real-time weight adaptive strategy further includes: when the window variance of any data in the three types of vector data is greater than the corresponding fluctuation threshold, the current fluctuation of that type of data is abnormal, and the real-time reliability is reduced to the base reliability. ,in, .

[0106] Specifically, when the window variance of a certain type of deviation vector exceeds the corresponding fluctuation threshold, the current fluctuation of that type of data is abnormal, and the real-time reliability decreases from the basic reliability. The formula is: , in, For the first Real-time reliability of class bias; For the first The basic confidence level of the type of bias is a preset fixed value; The confidence decay coefficient is a positive integer, determined by the magnitude by which the window variance exceeds the fluctuation threshold. The window variance exceeds the amplitude coefficient; It is a positive integer. .

[0107] The real-time weight adaptive strategy also includes: normalizing the real-time reliability of the three types of vector data to obtain the real-time weights of the three types of vector data.

[0108] Specifically, real-time reliability directly corresponds to the initial weights of various biases. To ensure that the sum of the real-time weights of the three types of biases is 1, the real-time reliability of the three types of biases needs to be normalized to obtain the real-time weights of the three types of vector data. The formula for calculating the real-time weights is: , , , in, , , These are the real-time weights for geometric parameter deviation, accuracy parameter deviation, and dynamic state deviation, respectively, with values ​​ranging from [0,1], and satisfying the following conditions: ; , , The real-time reliability of the three types of biases is shown below.

[0109] In this embodiment, model deviation analysis and real-time weight adaptive strategy provide deviation judgment and weight support for dynamic correction, further improving the rationality and pertinence of the correction. The ICP optimization algorithm is used to align the two models, and overlap verification ensures a unified comparison benchmark, effectively avoiding analytical inaccuracies caused by alignment deviations. Three types of deviation vectors are extracted and normalized to eliminate dimensional differences, providing a reliable foundation for subsequent weighted calculations. The real-time weight adaptive strategy uses window variance analysis to analyze deviation fluctuations and dynamically adjusts the real-time reliability and weight of various deviations, ensuring that weight allocation aligns with the real-time processing status, balancing the degree of deviation impact and data reliability. The weighted calculation of comprehensive deviation factors quantifies the overall processing deviation, intuitively reflecting the degree to which the processing status deviates from the preset standard. The method also includes: when the comprehensive deviation factor is greater than the deviation threshold, taking the processing plan of the screw body to be processed and the preset model of the processed body as the correction target, and dynamically correcting the model comparison deviation through the deviation compensation strategy to obtain the dynamic correction instruction.

[0110] The screw body to be machined is dynamically corrected according to the dynamic correction command until the comprehensive deviation factor is less than the deviation threshold, then the dynamic correction of the screw machining is completed.

[0111] Furthermore, the deviation compensation strategy includes: prioritizing the three types of deviations according to their real-time weights, with the deviation having the higher real-time weight being compensated first.

[0112] Specifically, the real-time weights output based on the real-time weight adaptive strategy. , , The three types of deviations are prioritized, with the logic being that the higher the real-time weight, the greater the impact of the deviation on the processing quality, and the higher the compensation priority.

[0113] First, compare the real-time weights of the three types of biases, sort them from largest to smallest, and obtain the compensation priority order, such as... The compensation priority is as follows: the deviation of the precision parameter is greater than the deviation of the geometric parameter, which is greater than the deviation of the dynamic state.

[0114] If the real-time weights of two or three types of deviations are equal, they are sorted according to a preset priority rule: in high-precision screw machining, the priority of precision parameter deviation is greater than that of geometric parameter deviation, which is greater than that of dynamic state deviation; in ordinary precision screw machining, the priority of geometric parameter deviation is greater than that of precision parameter deviation, which is greater than that of dynamic state deviation.

[0115] For geometric parameter deviations, a geometric deviation compensation algorithm is used, with the design parameters of the screw body to be processed as the correction target, to calculate the corresponding geometric correction vector.

[0116] Specifically, the geometric parameter deviation is based on the set of design parameters of the screw to be processed. Using this as a benchmark, ensure that the geometric parameters of the real-time model after compensation are consistent with the design parameters, and eliminate geometric parameter deviations.

[0117] A geometric correction algorithm based on the least squares method is adopted, combined with the geometric parameter deviation vector. Calculate the geometric correction vector The formula is: , in, For geometric correction vectors, For the first The correction amount for each geometric parameter; For the actual model of the real-time machining body, the first The current values ​​of the geometric parameters; The first parameter in the design parameters of the screw to be processed Standard values ​​of each geometric parameter; The number of geometric parameters; To find the correction vector that minimizes the objective function This ensures that the deviation of the corrected geometric parameters is minimized.

[0118] For the deviation of accuracy parameters, the accuracy compensation vector is calculated by using the accuracy parameter deviation compensation algorithm and taking the process requirements of the screw body to be processed as the correction target.

[0119] Specifically, the accuracy parameter deviation is based on the set of process requirements for the screw to be processed. Based on this, ensure that the accuracy parameters of the real-time model after compensation meet the process requirements and eliminate accuracy parameter deviations.

[0120] An accuracy compensation algorithm based on error tracing is adopted, combined with the normalized deviation value of the accuracy parameter. Calculate the accuracy compensation vector The formula is: , , in, For the first The compensation amount for each precision parameter; For the first The maximum permissible deviation value for each accuracy parameter; For the first Normalized deviation values ​​of each precision parameter; For a sign function, when hour, ;when hour, ;when hour, ; This refers to the number of precision parameters.

[0121] For dynamic state deviations, a dynamic compensation algorithm is used to calculate the real-time dynamic correction vector with the pre-set model of the screw body to be processed as the correction target.

[0122] Specifically, the dynamic state deviation is based on the pre-set model of the processed body. Dynamic state parameters in Based on this, ensure that the dynamic state parameters of the real-time model after compensation are consistent with the preset model, stabilize the processing, and avoid the accumulation of deviations caused by dynamic fluctuations; A dynamic compensation algorithm based on PID control is adopted, combined with dynamic state deviation vector. Calculate the real-time dynamic correction vector The formula is: , , in, for Time of the first Real-time correction of each dynamic state parameter; This is the proportionality coefficient; The integral coefficient; These are the differential coefficients; for The deviation value of the m-th dynamic state parameter at time m; From the beginning of processing to At that moment, the The integral value of the deviation of each dynamic state parameter; for Time of the first The rate of change of the deviation of each dynamic state parameter; This represents the number of dynamic state parameters.

[0123] The geometric correction vector, the precision compensation vector, and the dynamic correction vector constitute the dynamic correction command.

[0124] Specifically, the qualified geometric correction vector Precision compensation vector and real-time dynamic correction vector Integrate and combine processing steps. timestamp and tool number We construct structured dynamic correction instructions, whose format includes core information such as compensation priority, correction parameter type, correction amount, correction timing, and execution device, to ensure that the processing equipment can recognize and execute correction operations.

[0125] The screw body to be machined is dynamically corrected according to the dynamic correction command until the comprehensive deviation factor is less than the deviation threshold, then the dynamic correction of the screw machining is completed.

[0126] Specifically, after the dynamic correction command is executed, real-time monitoring data is re-acquired, and the feature extraction, modeling, and deviation analysis processes are repeated to calculate a new comprehensive deviation factor. ,judge If the deviation factor is less than the deviation threshold, the deviation compensation is effective and the dynamic correction is completed. If the deviation factor is greater than the threshold, the compensation is not up to standard. The compensation vector, such as the correction amount and compensation priority, is readjusted and the compensation operation is performed again. This process is repeated until the comprehensive deviation factor is less than the deviation threshold, thus achieving dynamic correction of the screw machining.

[0127] In this embodiment, the deviation compensation strategy and dynamic correction closed-loop process achieve efficient correction of machining deviations, strengthening the screw machining quality defense line. Based on real-time weighted ranking of compensation priorities, and combined with the screw precision level, differentiated ranking rules are formulated to ensure that deviations with the greatest impact on machining quality are corrected first. Adaptive compensation algorithms are used for the three types of deviations to calculate correction vectors, ensuring that the compensation amount is scientifically reasonable and effectively eliminates various deviations. Structured dynamic correction instructions can be recognized and executed by the machining equipment, and the closed-loop correction mechanism continuously iterates and optimizes until the comprehensive deviation factor reaches the standard, completely solving the problem of machining deviation accumulation and significantly improving the screw machining accuracy and quality consistency.

[0128] Example 2 like Figures 3-6 As shown, a real-time dynamic correction system for screw machining is applied to the real-time dynamic correction method for screw machining described in Embodiment 1 above. The method includes: a screw clamping mechanism 1, a tool spindle 2, a tool magazine assembly 3, a tool changer 4, a sensor tool 5, and a control center 6.

[0129] The screw clamping mechanism 1 is set at both ends of the screw body to be processed. It is used to clamp, position and fix the screw body to be processed, ensuring that there is no radial runout and axial displacement of the screw body during the processing, providing a stable workpiece reference for screw processing; and driving the screw body to be processed to perform high-speed rotational motion, providing power support for screw cutting; and adjusting the speed and rotation accuracy in real time according to the dynamic correction command issued by the control center.

[0130] The tool spindle 2 is located above the machining area of ​​the system. The tool is clamped below the tool spindle 2, and the screw body to be machined is turned and milled by the drive mechanism.

[0131] The tool magazine assembly 3 is located on one side of the tool spindle and is used to store various tools and sensor tools 5 required for machining the screw. It classifies, stores and manages the tools and provides tool position information for the tool changing actuator.

[0132] The tool changing actuator 4 is located between the tool spindle 2 and the tool magazine assembly 3. It is used to grab the corresponding tool from the tool magazine assembly according to the instructions of the control center and transfer it to the tool spindle 2 to complete the automatic tool changing.

[0133] The tool changing actuator 4 includes a tool changing arm 401 and a drive assembly 402, wherein the tool changing arm 401 is mounted on the drive end of the drive assembly 402.

[0134] The tool changer arm 401 is used to grab the tool and sensor tool 5 in the tool magazine assembly and dock with the tool spindle 2.

[0135] The drive assembly 402 is used to drive the tool changer arm 401 to complete rotation and extension movements.

[0136] The sensor tool 5 is equipped with a monitoring module, which is used to dynamically collect raw monitoring data during the machining process of the screw body to be processed, and to timestamp each raw monitoring data.

[0137] The control center 6 is electrically connected to the screw clamping mechanism 1, the tool spindle 2, the tool magazine assembly 3, the tool changer 4, and the sensor tool, and is used to execute all the strategies in the real-time dynamic correction method for screw machining.

[0138] Specifically, during system operation, all components work together, and the control center 6 coordinates and manages the process according to the real-time dynamic correction method for screw machining. First, the screw clamping mechanism 1 clamps, positions, and fixes the screw body to be machined, ensuring that it is coaxially aligned with the tool spindle 2 without radial runout or axial displacement, while simultaneously driving the screw to rotate, providing power support for cutting. Subsequently, the control center 6 instructs the tool magazine assembly 3 to retrieve the required tool, and the tool changing actuator 4 grabs the tool through the tool changing arm 401. Driven by the drive assembly 402, it completes rotation and extension movements, transferring the tool to the tool spindle 2 and completing the docking. The tool spindle 2 drives the tool to move, performing milling and turning on the high-speed rotating screw. During the machining process, the monitoring module of the sensor tool 5 synchronously collects raw monitoring data such as cutting force and temperature, marks them with timestamps, and transmits them to the control center 6 in real time. The control center 6 performs data preprocessing, feature extraction, real-time model construction, deviation analysis, and weighted adaptive strategies. When the comprehensive deviation factor exceeds the threshold, a dynamic correction command is generated and sent to the screw clamping mechanism 1 to adjust the speed and rotation accuracy, and to the tool spindle 2 to adjust the machining parameters, etc., to complete the deviation compensation. After correction, the system repeats the monitoring and analysis process, iterating until the comprehensive deviation factor reaches the standard. After machining is completed, the tool changer 4 returns the tool to the tool magazine assembly 3, and the screw clamping mechanism 1 releases the screw, completing the entire machining process.

[0139] In this embodiment, the positioning and rotation control of the screw clamping mechanism 1 ensures the stability of the machining datum from the workpiece end, avoiding machining quality problems caused by clamping deviations and speed fluctuations, and laying the foundation for correction. The automated tool changing design of the tool magazine assembly 3 and the tool changing actuator 4 reduces manual intervention, improves machining efficiency, and at the same time realizes standardized tool management, avoiding misuse and mixing of tools, and ensuring machining consistency.

[0140] The real-time monitoring function of sensor tool 5 provides high-quality raw data for the correction method, ensuring the accuracy of subsequent data preprocessing and feature extraction, and achieving full traceability and monitoring of the machining process. The core role of control center 6 in overall management and control enables the execution of all strategies in the correction method and the coordinated linkage of various components, ensuring rapid response and implementation of dynamic correction commands. The entire system deeply integrates the mechanical structure with the correction method, improving machining efficiency and automation level, and effectively eliminating various machining deviations through a closed-loop correction mechanism. This significantly improves the machining accuracy and quality stability of the screw, adapts to the machining needs of screws with different accuracy levels, and reduces manual operation intensity and machining scrap rate. like Figure 2 As shown, the control center 6 further includes a pre-processing input module 601, which is used to receive the processing pre-processing plan of the screw body to be processed, and to parse and store the design parameters and process requirements in the processing pre-processing plan.

[0141] Specifically, the contingency plan input module 601 can not only analyze the core design parameters and process requirements in the contingency plan, but also classify and store the analyzed parameters to establish a dedicated parameter file. It also has a contingency plan verification function, which can quickly identify problems such as missing parameters and logical contradictions in the contingency plan, and promptly provide feedback to staff for supplementation and improvement, providing complete and reliable basic data support for the subsequent construction of preset models.

[0142] The calling module 602 is electrically connected to the pre-plan input module. The calling module 602 has a built-in preset screw processing parameter database, a historical database of similar screw processing process parameter cases, and a basic model template of the screw processing body. The calling module 602 is used to execute the calling strategy, classify and analyze the design parameters and process requirements of the screw body to be processed, match process parameters, adapt and adjust the basic model and calibrate errors, and output the preset model of the processing body.

[0143] In this embodiment, the invocation strategy first classifies and parses the received design parameters and process requirements, distinguishing core categories such as geometric parameters, precision parameters, and process parameters. Then, based on the precision level of the specifications of the screw to be processed, it matches the case with the highest similarity from the historical case library to select suitable process parameters. Subsequently, it calls the basic model template, adapts and adjusts it according to the parsed design parameters, and generates an initial preset model. Finally, it uses an error calibration algorithm to calibrate the initial model in combination with historical case deviation data, eliminating deviations generated during template adaptation, and ultimately outputting a preset model of the processed body with the required precision.

[0144] The monitoring module 603 includes a sensor on the cutting tool. The monitoring module is used to dynamically collect raw monitoring data during the machining process of the screw body to be processed, and add a timestamp mark to each raw monitoring data.

[0145] Specifically, the monitoring module can collect various raw monitoring data during the machining process, covering core indicators such as cutting force, cutting temperature, vibration frequency, spindle speed, feed rate, surface roughness, and contour deviation. During the acquisition process, a unique timestamp is added to each raw monitoring data point to ensure that each data point can be traced back to the specific machining time, machining process, and corresponding tool number. Simultaneously, the monitoring module has a real-time data transmission function, synchronously pushing the collected raw data to the data preprocessing module 604, and can monitor the data acquisition status in real time. If data interruption or acquisition abnormality occurs, a warning signal is promptly issued to remind personnel to check the sensors, tools, and transmission links.

[0146] Data preprocessing module 604 is used to execute data preprocessing strategies, employing... Criteria, normalization algorithms, and moving average filtering algorithms are used to screen outliers, normalize, and smooth the raw monitoring data. The processed effective real-time monitoring data are then classified, archived, and associated with corresponding machining processes, tool numbers, and design parameters to form a structured real-time monitoring dataset.

[0147] In this embodiment, the first step adopts The criteria involve outlier screening, calculating the arithmetic mean and standard deviation of various monitoring data, and identifying and removing outliers. The first step involves identifying abnormal data within a certain range and calculating the percentage of outliers. Based on this percentage, the stability of sensor acquisition is assessed, and calibration prompts are issued and missing data is supplemented if necessary. The second step employs a normalization algorithm to uniformly map the screened valid data to the [0,1] interval, eliminating the weight imbalance caused by different units and numerical ranges. At the same time, the normalization results are verified to ensure that all data meet the interval requirements. The third step uses a moving average filtering algorithm to adjust the size of the filtering window according to the screw machining accuracy, filtering out high-frequency noise in the data and restoring the true trend of the machining state.

[0148] The feature extraction module 605 is electrically connected to the data preprocessing module 604. The feature extraction module 605 is used to execute the feature extraction strategy, classify the structured real-time monitoring dataset, extract basic features, screen key features, quantize and calibrate it, and output the real-time state feature set of the screw body to be processed.

[0149] Specifically, the feature extraction module 605 classifies the dataset into three categories based on data attributes and their correlation with processing quality: processing status features, equipment operation features, and workpiece inspection features. Then, for each category, it extracts corresponding basic statistical features and core indicators, such as the peak and fluctuation amplitude of cutting force in processing status features, the spindle speed stability in equipment operation features, and the average surface roughness in workpiece inspection features, and completes quantification calculations using corresponding formulas. Next, it calls a preset feature importance evaluation model, combines screw processing quality indicators, and uses the Pearson correlation coefficient to quantify the correlation between each feature and processing quality, selecting key features with correlation coefficients higher than a preset threshold and eliminating redundant features. Finally, it quantifies the key features, mapping them uniformly to the [0,10] interval. Simultaneously, it combines the design parameters and process requirements of the screw to be processed, calibrating the normal range and abnormal warning threshold of each key feature. After feature verification, it outputs a real-time status feature set of the screw body to be processed, providing feature input for real-time modeling.

[0150] The modeling module 606 is electrically connected to the calling module 602 and the feature extraction module 605 respectively. The modeling module 606 is used to execute the modeling strategy, receive the pre-set model of the workpiece output by the calling module and the real-time state feature set output by the feature extraction module, render and fuse the real-time state feature set into the pre-set model of the workpiece, and construct and output the real-time actual model of the workpiece.

[0151] Specifically, based on the machining process nodes and timestamps in the real-time state feature set, a correspondence is established between real-time features and preset model feature nodes. This ensures that machining state features such as cutting force and cutting temperature correspond to the cutting area nodes of the preset model, equipment operation features such as spindle speed correspond to process parameter association nodes, and workpiece detection features such as surface roughness correspond to accuracy detection nodes, avoiding misalignment between features and model nodes. Subsequently, a weighted fusion algorithm is used to calculate in segments according to the machining process. Fusion weights are assigned according to the correlation between features and model nodes, and the quantified values ​​of real-time state features are dynamically rendered and fused to the corresponding nodes of the preset model, realizing dynamic model updates. After fusion, the model is verified to ensure that it can truly reflect the real-time machining state. Finally, a real-time workpiece actual model is constructed and output.

[0152] The deviation analysis module 607 is electrically connected to the calling module 602 and the modeling module 606 respectively. The deviation analysis module 607 is used to execute the model deviation analysis strategy and the real-time weight adaptive strategy. It calls the model feature alignment algorithm to realize the alignment of the real-time workpiece actual model with the preset model of the machining body, extracts various deviation vectors and performs normalization processing, calculates the real-time weight of various deviations through the real-time weight adaptive strategy, and then obtains the model comparison deviation and the comprehensive deviation factor through weighted calculation. At the same time, it presets the deviation threshold and completes the comparison between the comprehensive deviation factor and the deviation threshold.

[0153] Specifically, firstly, a model feature alignment algorithm based on ICP optimization is invoked to align the real-time workpiece model output by the modeling module with the preset machining model output by the calling module, focusing on aligning key feature nodes, geometric contours, and accuracy benchmarks. The model overlap is calculated and verified to ensure alignment is qualified. Next, the geometric parameter deviation vector, accuracy parameter deviation vector, and dynamic state deviation vector of the two models are extracted, and the three types of deviation vectors are normalized to eliminate dimensional differences. Then, a real-time weight adaptive strategy is executed, setting a preset time window, calculating the window variance of the three types of deviation vectors, comparing it with the corresponding fluctuation threshold, adjusting the real-time reliability of each type of deviation, and then obtaining the real-time weights of each type of deviation through normalization. Finally, the model comparison deviation and comprehensive deviation factor are obtained through weighted calculation, while preset deviation thresholds adapted to different precision screws are used to complete the comparison between the comprehensive deviation factor and the deviation threshold. The compensation module 608 is electrically connected to the deviation analysis module 607. The compensation module 608 is used to execute a deviation compensation strategy. According to the real-time weights output by the deviation analysis module, the three types of deviations are prioritized. The geometric correction vector, the accuracy compensation vector, and the real-time dynamic correction vector are calculated by the corresponding compensation algorithm. The three types of vectors are integrated to form a dynamic correction instruction.

[0154] Specifically, firstly, the three types of deviations are prioritized based on real-time weights, with higher real-time weights resulting in higher compensation priority. If the weights are equal, a preset sorting rule is applied based on the screw's accuracy level. Then, appropriate compensation algorithms are used for each type of deviation. For geometric parameter deviations, a least-squares-based geometric correction algorithm is employed, using the design parameters of the screw to be processed as a benchmark to calculate the geometric correction vector, ensuring that the compensated geometric parameters are consistent with the design parameters. For accuracy parameter deviations, an error-based accuracy compensation algorithm is used, combining the normalized deviation value to calculate the accuracy compensation vector, ensuring that the compensated accuracy parameters meet process requirements. For dynamic state deviations, a PID-based dynamic compensation algorithm is used, combining the deviation change rate and integral value to calculate the real-time dynamic correction vector, stabilizing the machining process. Finally, the three types of correction vectors are integrated, associated with machining process nodes, timestamps, and tool numbers, to construct a structured dynamic correction instruction containing core information such as compensation priority, correction amount, and correction timing.

[0155] The execution module 609 is electrically connected to the compensation module 608 and the monitoring module 603 respectively. The execution module 609 is used to receive the dynamic correction command output by the compensation module, control the processing equipment to perform dynamic processing correction on the screw body to be processed; and receive the correction process monitoring data fed back by the monitoring and preprocessing module in real time, and cooperate with the deviation analysis module to cyclically calculate the comprehensive deviation factor until the comprehensive deviation factor is less than the deviation threshold, and then stop the correction operation.

[0156] Specifically, on the one hand, the system receives dynamic correction commands output by the compensation module 608, parses the core information in the commands, and sends control signals to the machining equipment, such as the screw clamping mechanism 1 adjusting its rotation speed and rotational accuracy, and the tool spindle 2 adjusting its cutting speed and feed rate, to control the equipment to perform dynamic correction operations. On the other hand, the system receives real-time monitoring data of the correction process from the monitoring module 603 and pushes it to the data preprocessing module 604. Together with the deviation analysis module 607, the system cyclically executes the data preprocessing, feature extraction, modeling, and deviation analysis process, recalculates the comprehensive deviation factor, and stops the correction operation immediately when the comprehensive deviation factor is less than the deviation threshold, thus completing the dynamic correction and ensuring that the screw machining accuracy meets the standards.

[0157] In this embodiment, the pre-processing input module 601 ensures the integrity and reliability of input data by parsing processing pre-processing plans, classifying and storing parameters, and performing pre-processing verification functions. This avoids problems such as missing parameters and logical contradictions from the source, providing solid data support for subsequent pre-set model construction and ensuring the uniformity of processing benchmarks. The invocation module 602 relies on the built-in database and templates to achieve process parameter matching and model adaptation calibration through invocation strategies. This solves the problem of poor reusability of parameters for similar screws, improves the accuracy of pre-set models, and provides a reliable benchmark for deviation analysis.

[0158] The monitoring module 603 enables real-time monitoring of the processing, comprehensively collecting various core indicators and adding timestamps to ensure data traceability. It also features anomaly warning functionality to guarantee the continuity and effectiveness of data collection. The data preprocessing module 604 optimizes the data through a combination of multiple algorithms, effectively removing outliers, filtering noise, and eliminating dimensional differences to form a structured dataset. This provides high-quality data support for feature extraction and avoids data distortion that could interfere with subsequent analysis.

[0159] The feature extraction module 605 classifies features and filters key indicators, eliminates redundant information, and quantifies and calibrates the feature range, improving the efficiency and relevance of subsequent modeling and deviation analysis. The modeling module 606 integrates real-time features with a preset model to construct a real-time model that accurately reflects the processing status, providing a reference for deviation analysis. The deviation analysis module 607 quantifies the overall processing deviation through alignment and scientific weight calculation, providing a basis for compensation strategies. The compensation module 608 prioritizes deviations and uses an adaptation algorithm to calculate correction vectors, ensuring efficient compensation. The execution module 609 executes correction instructions and implements a closed-loop cycle, ensuring continuous optimization of processing deviations until the target is met.

[0160] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the claims of this invention should still fall within the patent coverage of this invention.

Claims

1. A real-time dynamic correction method for screw machining, characterized in that, The method includes: Input the machining plan for the screw body to be machined, which includes the design parameters and process requirements of the screw body to be machined; By invoking strategies to analyze the design parameters and process requirements of the screw body to be processed, a preset model of the processed body is obtained; The machining process of the screw body to be machined is dynamically monitored by the monitoring module on the sensor tool, and real-time monitoring data is obtained through data preprocessing strategy. By analyzing real-time monitoring data using feature extraction strategies, the real-time state characteristics of the screw body to be processed can be obtained. Real-time state characteristics are used to construct an actual model of the real-time processing body through modeling strategies; The model deviation analysis strategy is used to compare the real-time workpiece actual model with the workpiece preset model to obtain the model comparison deviation and comprehensive deviation factor. When the comprehensive deviation factor is greater than the deviation threshold, the machining plan of the screw body to be processed and the preset model of the machined body are used as the correction targets. The deviation of the model comparison is dynamically corrected through the deviation compensation strategy to obtain the dynamic correction command. The screw body to be machined is dynamically corrected according to the dynamic correction command until the comprehensive deviation factor is less than the deviation threshold, then the dynamic correction of the screw machining is completed.

2. The real-time dynamic correction method for screw machining according to claim 1, characterized in that, The invocation strategy includes: The system calls up a pre-set screw machining parameter database, classifies and analyzes the design parameters of the screw body to be machined, and archives the analyzed parameters according to geometric parameters and precision parameters. Based on the process requirements in the processing plan, the process parameter case library of similar screw processing in the past is called, and the standard process requirements that match the design parameters and material properties of the screw to be processed in the current process with a greater than the preset proportion are selected. The process parameters in the standard process requirements are then extracted as standard parameters. Call the preset screw machining body basic model template, and adjust the parameters of the basic model template according to the analyzed geometric parameters, accuracy parameters and standard parameters to clarify the key feature nodes in the model and the accuracy benchmark of the model; After calling the error calibration algorithm to make preliminary corrections to the geometric and process deviations in the basic model template, the preset model of the processed body is obtained.

3. The real-time dynamic correction method for screw machining according to claim 1, characterized in that, The data preprocessing strategy includes: The raw monitoring data during the machining process is collected synchronously by the monitoring module on the sensor tool, and each raw monitoring data is timestamped. use The criteria, normalization algorithm, and moving average filtering algorithm are used to screen outliers, normalize, and smooth the collected raw monitoring data to obtain effective real-time monitoring data. The preprocessed real-time monitoring data is classified and archived, and associated with the corresponding machining process, tool number and design parameters to form a structured real-time monitoring dataset.

4. The real-time dynamic correction method for screw machining according to claim 1, characterized in that, The feature extraction strategy includes: The real-time monitoring dataset is classified into processing status features and equipment operation features, and basic features of each type of data are extracted. The processing status features include the peak value and fluctuation amplitude of cutting force, the steady-state value and rate of change of cutting temperature, the peak value of vibration frequency and the distribution of the main frequency. The equipment operation features include the stability of spindle speed and the fluctuation of feed rate. The workpiece detection features include the average surface roughness of screw machining and the initial value of contour deviation. At the same time, the timestamps and processing operation nodes corresponding to each feature are associated to form an initial feature set. The preset feature importance evaluation model is invoked, and the correlation coefficient between each feature in the initial feature set and the screw processing quality is calculated in combination with the precision requirements and process characteristics of screw processing. Key features with correlation coefficients greater than the preset threshold are then selected. The key features after screening are quantified, the quantification standards and threshold ranges of each feature are set, and the design parameters and process requirements of the screw to be processed are called to calibrate each key feature to obtain the normal range of the feature and the abnormal warning threshold. The key features after quantitative calibration are integrated to obtain the real-time state feature set of the screw body to be processed.

5. The real-time dynamic correction method for screw machining according to claim 1, characterized in that, The modeling strategy includes: The real-time state feature set of the screw body to be processed is rendered and fused into the preset model of the processing body to obtain the actual model of the real-time workpiece.

6. The real-time dynamic correction method for screw machining according to claim 1, characterized in that, The model bias analysis strategy includes: The model feature alignment algorithm is called to align the key feature nodes, geometric contours, and accuracy benchmarks of the real-time workpiece actual model with the preset model of the machining body. The geometric parameter deviation vector, accuracy parameter deviation vector, and dynamic state deviation vector of the two models are extracted and normalized to obtain the corresponding normalized deviation values. The real-time weights of various deviations are obtained through a real-time weight adaptive strategy, and then weighted with the corresponding normalized deviation values ​​to obtain a comprehensive deviation factor.

7. The real-time dynamic correction method for screw machining according to claim 6, characterized in that, The real-time weight adaptive strategy includes: calculating the window variance of three types of vector data within a preset window: geometric parameter deviation vector, precision parameter deviation vector, and dynamic state deviation vector, and comparing them with the corresponding fluctuation thresholds; Calculate the real-time reliability of the three types of vector data based on the window variance of the three types of vector data; When the window variance of any data point in the three types of vector data exceeds the corresponding fluctuation threshold, the current fluctuation of that data type is abnormal, and the real-time reliability is reduced to the base reliability. ,in, ; The real-time reliability of the three types of vector data is normalized to obtain the real-time weights of the three types of vector data.

8. The real-time dynamic correction method for screw machining according to claim 1, characterized in that, The deviation compensation strategy includes: prioritizing the three types of deviations according to their real-time weights, with the deviation having the higher real-time weight being compensated first. For geometric parameter deviations, a geometric deviation compensation algorithm is used, with the design parameters of the screw body to be processed as the correction target, to calculate the corresponding geometric correction vector; For the deviation of accuracy parameters, the accuracy compensation vector is calculated by using the accuracy parameter deviation compensation algorithm and taking the process requirements of the screw body to be processed as the correction target. For dynamic state deviations, a dynamic compensation algorithm is used to calculate the real-time dynamic correction vector with the pre-set model of the machining body of the screw body to be processed as the correction target; The geometric correction vector, the precision compensation vector, and the dynamic correction vector constitute the dynamic correction command.

9. A real-time dynamic correction system for screw machining, characterized in that, The system is applied to the real-time dynamic correction method for screw machining as described in any one of claims 1-8, the method comprising: a screw clamping mechanism (1), a tool spindle (2), a tool magazine assembly (3), a tool changer (4), a sensor tool (5), and a control center (6). The screw clamping mechanism (1) is set at both ends of the screw body to be processed. It is used to clamp, position and fix the screw body to be processed, so as to ensure that the screw body to be processed has no radial runout and axial displacement during the processing, and to provide a stable workpiece reference for screw processing; and to drive the screw body to be processed to perform high-speed rotational motion, providing power support for screw cutting; and to adjust the speed and rotation accuracy in real time according to the dynamic correction command issued by the control center. The tool spindle (2) is located above the machining area of ​​the system. The tool is clamped below the tool spindle (2) and the screw body to be machined is milled and turned by the drive of the moving drive mechanism. The tool magazine assembly (3) is located on one side of the tool spindle and is used to store various types of tools and sensor tools (5) required for machining the screw. It classifies, stores and manages the tools and provides tool position information for the tool changing actuator. The tool changing actuator (4) is located between the tool spindle (2) and the tool magazine assembly (3). It is used to grab the corresponding tool from the tool magazine assembly according to the instructions of the control center and transfer it to the tool spindle (2) to complete the automatic tool changing. The tool changing actuator (4) includes a tool changing arm (401) and a drive assembly (402), wherein the tool changing arm (401) is mounted on the drive end of the drive assembly (402); The tool changer arm (401) is used to grab the tool and sensor tool (5) in the tool magazine assembly and dock with the tool spindle (2); The drive assembly (402) is used to drive the tool changer arm (401) to complete rotation and extension movements; The sensor tool (5) is equipped with a monitoring module. The monitoring module is used to dynamically collect the original monitoring data during the machining process of the screw body to be processed, and to timestamp each original monitoring data. The control center (6) is electrically connected to the screw clamping mechanism (1), the tool spindle (2), the tool magazine assembly (3), the tool changer (4), and the sensor tool, and is used to execute all the strategies in the real-time dynamic correction method for screw machining.

10. The real-time dynamic correction system for screw machining according to claim 9, characterized in that, The control center (6) includes: The pre-processing input module (601) is used to receive the pre-processing plan of the screw body to be processed, and to parse and store the design parameters and process requirements in the pre-processing plan; The calling module (602) is electrically connected to the pre-plan input module. The calling module (602) has a built-in preset screw processing parameter database, a historical database of similar screw processing process parameter cases, and a basic model template of the screw processing body. The calling module (602) is used to execute the calling strategy, classify and analyze the design parameters and process requirements of the screw body to be processed, match process parameters, adapt and adjust the basic model and calibrate errors, and output the preset model of the processing body. (603), which is located on the sensor tool, the monitoring module (603) is used to dynamically collect the original monitoring data during the machining process of the screw body to be processed, and add a timestamp mark to each original monitoring data; The data preprocessing module (604) is used to execute data preprocessing strategies, employing... Criteria, normalization algorithms, and moving average filtering algorithms are used to screen outliers, normalize, and smooth the raw monitoring data. The processed effective real-time monitoring data are classified, archived, and associated with corresponding processing steps, tool numbers, and design parameters to form a structured real-time monitoring dataset. The feature extraction module (605) is electrically connected to the data preprocessing module (604); the feature extraction module (605) is used to execute the feature extraction strategy, classify the structured real-time monitoring dataset, extract basic features, screen key features, quantize and calibrate it, and output the real-time state feature set of the screw body to be processed; The modeling module (606) is electrically connected to the calling module (602) and the feature extraction module (605) respectively. The modeling module (606) is used to execute the modeling strategy, receive the pre-set model of the processing body output by the calling module and the real-time state feature set output by the feature extraction module, render and fuse the real-time state feature set into the pre-set model of the processing body, and construct and output the real-time workpiece actual model. The deviation analysis module (607) is electrically connected to the calling module (602) and the modeling module (606) respectively. The deviation analysis module (607) is used to execute the model deviation analysis strategy and the real-time weight adaptive strategy, call the model feature alignment algorithm to realize the alignment of the real-time workpiece actual model with the preset model of the machining body, extract various deviation vectors and perform normalization processing, calculate the real-time weight of various deviations through the real-time weight adaptive strategy, and then obtain the model comparison deviation and comprehensive deviation factor through weighted calculation. At the same time, the deviation threshold is preset and the comparison between the comprehensive deviation factor and the deviation threshold is completed. The compensation module (608) is electrically connected to the deviation analysis module (607). The compensation module (608) is used to execute a deviation compensation strategy. According to the real-time weight output by the deviation analysis module, the three types of deviations are prioritized and the geometric correction vector, accuracy compensation vector and real-time dynamic correction vector are calculated by the corresponding compensation algorithm. The three types of vectors are integrated to form a dynamic correction instruction. The execution module (609) is electrically connected to the compensation module (608) and the monitoring module (603) respectively. The execution module (609) is used to receive the dynamic correction command output by the compensation module, control the processing equipment to perform dynamic processing correction on the screw body to be processed; and receive the correction process monitoring data fed back by the monitoring and preprocessing module in real time, and cooperate with the deviation analysis module to calculate the comprehensive deviation factor in a loop until the comprehensive deviation factor is less than the deviation threshold, and then stop the correction operation.

Images