Intelligent anti-collision monitoring system and method applied to pipe numerical control machining

Abstract

The application discloses an intelligent anti-collision supervision system and method applied to pipe numerical control machining, relates to the technical field of industrial control software, and builds a multi-dimensional scene matching historical machining data set, and based on scene exclusive corner threshold fine segmentation track, extracts collision track core features, constructs a two-way association model of parameters and track features through two-dimensional cross verification, gets rid of the limitations of fixed parameters and general models, eliminates scene variable interference, accurately mines collision rules, and improves the accuracy of feature extraction and risk modeling. Through quantitative feature deviation, the collision risk is pre-judged, the dynamic correction parameters are deduced combined with the results, the cutting head running state is adaptively regulated and controlled, the database is updated synchronously, and the supervision closed loop is formed. It can not only avoid collision risk in advance, but also optimize the supervision accuracy through data iteration, solve the defects of existing technology such as prediction lag, regulation rigidity and insufficient scene adaptation.

Description

Technical Field

[0001] This invention relates to the field of industrial control software technology, specifically an intelligent anti-collision monitoring system and method applied to CNC machining of pipes. Background Technology

[0002] Collision prevention monitoring of the cutting head is a crucial step in CNC pipe machining. Existing technologies, when performing cutting head trajectory segmentation, collision feature analysis, and collision prediction, are affected by factors such as pipe material, cross-section type, and processing technology, leading to an inability to accurately determine collision tendencies in the non-collision stage of the predicted trajectory. They can only rely on adjusting the cutting head's operating state with fixed control parameters, ultimately resulting in insufficient scenario adaptability and accuracy of the cutting head collision prevention monitoring, failing to meet the collision prevention requirements of CNC pipe machining in different processing scenarios. Therefore, there is an urgent need for an intelligent collision prevention monitoring system and method for CNC pipe machining. Summary of the Invention

[0003] The purpose of this invention is to provide an intelligent anti-collision monitoring system and method for CNC machining of pipes, so as to solve the problems raised in the prior art.

[0004] To achieve the above objectives, the present invention provides the following technical solution: an intelligent anti-collision monitoring method applied to CNC machining of pipes, the intelligent anti-collision monitoring method comprising the following steps:

[0005] Acquire historical processing data that matches the real-time processing scenario, and construct a historical processing dataset with the same scenario. The historical processing dataset with the same scenario is a set of historical pipe CNC processing data that conforms to the consistency of the processing scenario after being filtered according to unified search parameters.

[0006] The process of constructing a historical processing dataset for the same scenario is as follows: set unified retrieval parameters, which include pipe material, cross-section type, processing technology type, cutting head model and CNC system type, and use the combination data of the unified retrieval parameters as a unique identifier for a processing scenario;

[0007] Based on the selection of unified search parameters, all matching data are extracted from the historical processing database. The matching data meets the requirements of having the same unified search parameters, no manual intervention in the processing process, and complete data records. The complete data records include cutting head posture, motion trajectory, collision situation, and control parameter information.

[0008] Integrate and match data to form a historical processing dataset of the same scenario, select a unified retrieval parameter as the key value, select matching data as the value, and integrate and store the data in the form of key-value pairs to obtain a set of historical CNC processing data for pipes.

[0009] By setting unified search parameters covering pipe material, cross-section type, processing technology type, cutting head model, and CNC system type, these parameters are combined as a unique identifier for a single processing scenario. Based on this, matching data with complete records and no human intervention are selected from the historical database. The dataset is then integrated and stored in key-value pairs to construct a historical processing dataset for the same scenario. This completely abandons the existing technology's generalized and scenario-indiscriminate historical data retrieval mode, achieving accurate matching between historical data and real-time processing scenarios from the data source end. This solidifies the data foundation for full-process scenario adaptation analysis and avoids the data analysis base distortion problem caused by different processing scenario factors.

[0010] The axial continuous motion trajectory of the cutting head in the historical CNC machining data set of pipes is segmented, and the feature data of each trajectory segment is extracted. The axial continuous motion trajectory of the cutting head represents the complete motion path of the cutting head along the pipe processing direction; the trajectory segment represents a continuous trajectory fragment divided according to the rotation angle threshold; the feature data includes the maximum lateral coverage of the cutting head, the length of the trajectory segment, the curvature of the trajectory segment, and the distance the trajectory segment is drawn when a collision occurs.

[0011] The process of setting the corner threshold involves separating the corner data corresponding to the trajectory segments that have collided and the corner data corresponding to the trajectory segments that have not collided from the historical CNC machining data set of pipes.

[0012] Count all the turning angle values ​​of the trajectory segments that have collided, extract the minimum value, and record it as the minimum collision angle.

[0013] Count all the turning angle values ​​of the trajectory segments that did not collide, extract the maximum value, and record it as the maximum non-collision turning angle;

[0014] Calculate the arithmetic mean of the minimum collision angle and the maximum non-collision angle, and determine the arithmetic mean as the initial angle threshold;

[0015] A predetermined proportion of trajectory data is randomly selected from the historical CNC machining data set of pipes as verification samples, and the trajectory of the verification samples is segmented according to the initial turning angle threshold.

[0016] In the statistical verification sample, the degree of overlap between the trajectory segments divided according to the initial turning angle threshold and the historically actual trajectory segments;

[0017] If the overlap does not meet the preset standard, the initial corner threshold is adjusted by the preset step size, and the verification process is repeated until the overlap meets the preset standard. The final adjusted value is then determined as the corner threshold.

[0018] The segmentation process of the axial continuous motion trajectory of the cutting head is to check the angle value of each sampling point on the trajectory point by point, mark the sampling point whose angle value exceeds the angle threshold as the segment point, and divide the trajectory into multiple continuous small segments according to the interval of adjacent segment points.

[0019] The calculation of the maximum lateral coverage of the cutting head is based on the vertical plane of the direction of movement of the processing surface, which is perpendicular to the axis of pipe processing.

[0020] The length of a trajectory segment is calculated by obtaining the three-dimensional spatial coordinates of the starting and ending points of the trajectory segment, and then calculating the distance between the two points based on the spatial straight-line distance formula.

[0021] The curvature of a small segment of the trajectory is calculated by collecting the three-dimensional spatial coordinates of all sampling points on the small segment of the trajectory, and then calculating the curvature value point by point based on the curvature calculation formula and taking the arithmetic mean.

[0022] The distance drawn by the trajectory segment during the collision is extracted to locate the three-dimensional spatial coordinates of the collision start point and the collision end point in the trajectory segment, and the trajectory arc length between the two points is calculated.

[0023] The extraction process of all feature data is based on the small trajectory segments that have collided in the historical CNC machining data set of pipe materials. The extraction results are used as training samples for the association mapping model.

[0024] To address the trajectory characteristic variations caused by differences in processing scenarios, this method separates the angle values ​​of collision and non-collision trajectories from historical data, iteratively calculates and verifies a specific angle threshold to replace the fixed angle parameters of existing technologies. Then, it verifies the cutting head trajectory sampling points point by point according to this specific threshold, accurately dividing continuous trajectory segments. At the same time, it calculates various trajectory feature data according to scenario-based benchmarks, extracting effective feature samples only based on collision trajectories. This aligns with the actual trajectory operation rules of the corresponding processing scenario, eliminating the interference of scenario factors such as pipe material and processing technology on trajectory segmentation and feature extraction, and ensuring that the extracted feature data is fully adapted to the current processing scenario.

[0025] Based on the analysis of the control parameters and trajectory segment features in historical data, a two-way correlation mapping relationship is obtained, forming a correlation mapping model between parameters and trajectory segment features. The control parameters include the cutting head attitude angle and motion speed.

[0026] The process of segmenting the continuous axial motion trajectory of the cutting head and extracting feature data to form an association mapping model includes:

[0027] A turning angle threshold is set as the segmentation criterion. When the turning angle of any point on the continuous axial movement trajectory of the cutting head exceeds the turning angle threshold, that point is the segmentation point. The complete trajectory is divided into multiple continuous trajectory segments based on the segmentation point.

[0028] For each trajectory segment, the maximum lateral coverage of the cutting head is obtained by calculating the angle between the vector from the starting point to the ending point of the trajectory segment and the set running direction vector. The length of the trajectory segment is obtained by calculating the straight-line distance between the starting point and the ending point. The curvature of the trajectory segment is obtained by calculating the mean curvature of all points on the trajectory segment. For trajectory segments that collide, the distance drawn by the trajectory segment at the time of collision is obtained by extracting the trajectory length from the collision starting point to the collision ending point. Based on the two selection results, the correlation mapping model is analyzed through cross-validation.

[0029] The first screening result is to screen historical data segments with the same values ​​across all dimensions of the control parameters, compare and analyze the corresponding trajectory segment feature data, and statistically obtain the overlap interval of various trajectory segment feature data when the control parameters are fixed.

[0030] The second screening result is to screen historical data segments with the same short-segment feature data, extract the corresponding control parameters, and statistically obtain the value distribution of the control parameters when the trajectory is fixed.

[0031] Using a single trajectory segment as the smallest unit of analysis, the first screening result and the second screening result are cross-validated to establish a bidirectional correlation mapping relationship between the combination of control parameters and the feature data of the trajectory segment, and integrate them to form a correlation mapping model between parameters and the features of the trajectory segment.

[0032] The specific process of cross-validating the first and second screening results is as follows:

[0033] The first step is to extract any set of control parameter combinations and the overlapping intervals of all corresponding trajectory segment feature data from the first screening results, and mark the overlapping intervals of the control parameter combination and the corresponding trajectory segment feature data as the first association group.

[0034] The second step is to extract all control parameter value distributions that are consistent with the trajectory segment feature data in the first association group from the second screening results, and mark the trajectory segment feature data and the corresponding control parameter value distribution as the second association group.

[0035] The third step is to verify whether the combination of control parameters in the first association group falls completely within the distribution of control parameter values ​​in the second association group, and at the same time verify whether the trajectory segment feature data in the second association group falls completely within the overlapping range of trajectory segment feature data in the first association group.

[0036] Fourth, if both checks are satisfied, the association between the control parameter combination and the trajectory segment feature data is retained; if either check is not satisfied, the association is removed.

[0037] Fifth step: Traverse all first association groups and their corresponding second association groups, repeat the above verification process, and summarize all association relationships that pass the verification.

[0038] Using a single trajectory segment as the smallest unit of analysis, the verified correlation relationships are integrated to establish a bidirectional correlation mapping relationship between the combination of control parameters and the feature data of the trajectory segment, forming a correlation mapping model between parameters and the features of the trajectory segment.

[0039] By filtering historical data in two dimensions and conducting cross-validation, a bidirectional correlation mapping model between parameters and trajectory segment features is built. First, the values ​​of control parameters in all dimensions are fixed to statistically analyze the overlapping intervals of corresponding trajectory features. Then, the values ​​of trajectory features are fixed to extract the distribution of corresponding control parameters. Bidirectional verification is completed with a single trajectory segment as the smallest unit. This constructs a unique correlation relationship between control parameters and trajectory features that is fully adapted to the corresponding processing scenario, thus solving the shortcomings of existing technical models that lack scenario specificity and cannot adapt to different pipe processing conditions.

[0040] Based on the analysis of historical CNC machining data of pipes, the time required to predict the future trajectory of the cutting head is obtained, which is denoted as the prediction time. The motion data and control parameters of the cutting head that are currently running are collected and input into the trained machine learning model to predict the motion data of the cutting head within the prediction time.

[0041] Extract the running time span of all complete trajectories from the historical CNC machining data set of pipe materials. The running time span is the time difference between the start and end times of trajectory acquisition.

[0042] The extracted running time span is cleaned to remove outliers and retain the valid time span data. The arithmetic mean of the valid time span data is calculated and the result is used as the predicted duration of the future trajectory of the cutting head.

[0043] The training process of the machine learning model involves using the cutting head running data and control parameters of the collision trajectory that have occurred in the historical CNC machining data set of pipes as input samples, and the cutting head motion data within the corresponding time period as output samples to complete the training and verification of the model.

[0044] The motion data and control parameters of the cutting head during its current operating phase are collected. The motion data includes the three-dimensional spatial coordinates and attitude angles at each sampling time, and the control parameters include attitude angles and motion speed. These are input into a trained machine learning model, which outputs the cutting head motion data for the predicted duration.

[0045] Based on historical processing datasets from the same scenario, the complete trajectory running time span is cleaned and a specific prediction duration is calculated and determined. At the same time, machine learning model training is completed using scenario-specific historical collision trajectory data as samples. The motion data and control parameters of the cutting head currently running are collected in real time and input into the model. The cutting head motion data within the prediction duration is accurately output. The quantitative deduction of the trajectory in the non-collision stage is realized by relying on scenario-specific historical data, filling the gap in existing technologies that cannot perform trajectory prediction for specific processing scenarios and avoiding the deviation caused by generalized trajectory prediction logic.

[0046] Input the cutting head motion data within the predicted time period into the correlation mapping model, extract the feature information corresponding to the predicted motion data for analysis, and make predictions and supervision of the risk of processing collisions.

[0047] Extract the feature information corresponding to the cutting head motion data within the prediction time. The extraction logic is the same as the extraction method of the feature data of the historical collision trajectory segment. The feature information includes the maximum lateral coverage, length and curvature of the predicted trajectory segment.

[0048] The extracted feature information is input into the parameter and mapped to the trajectory segment feature association model, and the feature data of all collision trajectories in the model are retrieved.

[0049] Calculate the deviation between the predicted feature information and the collision feature data in the model. The deviation is the absolute value of the numerical difference between the feature data of each dimension.

[0050] The deviation value is compared with a preset feature deviation threshold. When the deviation value in any dimension is less than the feature deviation threshold, it is determined that the predicted trajectory segment has a collision tendency.

[0051] After the collision tendency is determined, a predictive regulatory conclusion is generated, which includes the location of the trajectory segment with collision tendency, the deviation value of the corresponding dimension, and the matching collision feature data.

[0052] The predicted trajectory features are compared with the scene-specific collision features in the associated mapping model by quantitative deviation comparison. The absolute value of the difference between the feature values ​​of each dimension is calculated. The predicted trajectory is judged to have a collision tendency by setting a preset feature deviation threshold. A predictive regulatory conclusion containing the specific trajectory location and deviation value is generated. By adopting a scenario-based feature comparison logic, potential collision risks in the non-collision stage are accurately identified, which greatly improves the pertinence and accuracy of collision avoidance prediction and realizes accurate identification of risks in advance.

[0053] Based on the results of predictive monitoring, dynamic correction parameters are derived and transmitted to the CNC machining system to automatically adjust the cutting head's operating status. At the same time, real-time machining data is fed back and stored in the historical pipe CNC machining data set.

[0054] Based on the control parameters corresponding to the collision feature data that match the collision tendency trajectory in the correlation mapping model, and combined with the deviation value in the predictive regulatory conclusion, the dynamic correction parameters are derived. The dynamic correction parameters include the cutting head attitude angle adjustment value and the motion speed adjustment value.

[0055] The dynamic correction parameters are transmitted to the control module of the CNC machining system. After receiving the parameters, the control module automatically adjusts the attitude angle and movement speed of the cutting head, and the adjustment direction is to increase the deviation value between the predicted feature and the collision feature.

[0056] The real-time processing data feedback storage process involves classifying the currently processed trajectory data, control parameters, dynamic correction parameters, and predictive regulatory conclusions according to unified retrieval parameters;

[0057] The categorized real-time processing data is added to the historical CNC machining data set for pipes, and the retrieval index information of the data set is updated after the addition is completed;

[0058] Based on the collision tendency prediction conclusion and the corresponding collision characteristics control parameters, and combined with the actual deviation value, scenario-specific dynamic correction parameters are derived to clarify the adjustment direction of the cutting head attitude and speed. The correction parameters are synchronously transmitted to the CNC system to complete automatic control, actively widening the characteristic deviation between the predicted trajectory and the collision trajectory to avoid collision risks. This abandons the rigid adjustment mode of fixed control parameters in existing technologies, realizes adaptive control of the cutting head's operating state under different processing scenarios, continuously improves the scenario adaptability and accuracy of anti-collision monitoring, and solves the problems of poor adaptability and inaccurate control in existing anti-collision monitoring technologies.

[0059] Furthermore, an intelligent anti-collision monitoring system for CNC machining of pipes includes a data set construction module, a trajectory feature processing module, a motion data prediction module, and a collision prediction and correction module.

[0060] The data set construction module is used to acquire historical processing data that matches the real-time processing scenario and construct a historical processing dataset that conforms to the consistency of the processing scenario. The trajectory feature processing module is used to segment the continuous axial motion trajectory of the cutting head in the historical pipe CNC processing dataset and extract the feature data of each trajectory segment. The motion data prediction module is used to determine the prediction duration based on the historical pipe CNC processing dataset and predict the cutting head motion data within the prediction duration through a machine learning model. The collision prediction and correction module is used to predict the processing collision tendency based on the correlation mapping model, derive dynamic correction parameters, and feed back and store the real-time processing data.

[0061] The data set construction module includes a retrieval parameter setting unit and a matching data integration unit. The retrieval parameter setting unit is used to set unified retrieval parameters and use the combined data of the unified retrieval parameters as a unique identifier for a type of processing scenario. The matching data integration unit is used to extract matching data based on the unified retrieval parameters and integrate it into a historical pipe CNC processing data set in the form of key-value pairs.

[0062] The trajectory feature processing module includes a trajectory segmentation execution unit and a feature data extraction unit. The trajectory segmentation execution unit is used to verify the turning angle value of each sampling point on the trajectory point by point, and to mark the sampling points that exceed the turning angle threshold as segmentation points, thus dividing the trajectory into continuous small segments. The feature data extraction unit is used to calculate and extract the maximum lateral coverage of the cutting head, the length of the small trajectory segment, the curvature of the small trajectory segment, and the distance the small trajectory segment is drawn when a collision occurs.

[0063] The motion data prediction module includes a prediction duration calculation unit and a motion data output unit. The prediction duration calculation unit is used to extract the running time span of historical trajectories, and calculates and determines the prediction duration of the future trajectory of the cutting head after data cleaning. The motion data output unit is used to train a machine learning model, collect the motion data and control parameters of the cutting head that have been running, input them into the model, and output the motion data of the cutting head within the prediction duration.

[0064] The collision prediction and correction module includes a collision tendency determination unit and a correction parameter feedback unit. The collision tendency determination unit is used to extract feature information from the predicted motion data, calculate the deviation value from the collision feature data, and compare it with a preset threshold to determine the collision tendency of the trajectory segment. The correction parameter feedback unit is used to derive dynamic correction parameters based on the corresponding control parameters in the associated mapping model and the deviation value, transmit them to the CNC machining system, and feed back and store real-time machining data.

[0065] Compared with the prior art, the beneficial effects of the present invention are:

[0066] 1. This invention establishes a historical processing dataset for multi-dimensional scenario matching, defines processing scenarios using unified retrieval parameters such as pipe material and processing technology, accurately filters compliant matching data, abandons the generalized data calling mode of existing technologies, avoids the analysis distortion problem caused by scenario differences from the data source end, makes trajectory analysis and risk prediction more in line with actual processing conditions, builds a solid data foundation for collision avoidance supervision, solves the core defect of poor data adaptability of existing technologies, and ensures the scenario-specificity of the whole process analysis.

[0067] 2. This invention completes fine-grained trajectory segmentation based on scene-specific corner thresholds, extracts core features of collision trajectories in a targeted manner, and constructs a bidirectional correlation model between control parameters and trajectory features through two-dimensional cross-validation. This breaks away from the limitations of fixed parameters and general models, effectively eliminates the interference of various scene variables on trajectory analysis, accurately mines collision patterns under corresponding working conditions, and significantly improves the accuracy of trajectory feature extraction and risk modeling, providing reliable dedicated model support for predicting collision tendencies in the non-collision stage.

[0068] 3. This invention completes the preliminary judgment of collision risk by quantifying feature deviation, and derives dynamic correction parameters based on the prediction results to achieve adaptive control of the cutting head's operating status. At the same time, it updates the historical database in real time to form a regulatory closed loop, which not only achieves the early avoidance of potential collision risks, but also continuously optimizes the regulatory accuracy through data iteration, solving the problems of lagging prediction, rigid control and insufficient adaptability of existing technologies. Attached Figure Description

[0069] Figure 1 This is a flowchart illustrating an intelligent anti-collision monitoring method for CNC machining of pipes according to the present invention. Detailed Implementation

[0070] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0071] Example 1: As Figure 1 As shown, the present invention provides a technical solution: an intelligent anti-collision monitoring method applied to CNC machining of pipes. The intelligent anti-collision monitoring method includes the following steps:

[0072] Acquire historical processing data that matches the real-time processing scenario, and construct a historical processing dataset with the same scenario. The historical processing dataset with the same scenario is a set of historical pipe CNC processing data that conforms to the consistency of the processing scenario after being filtered according to unified search parameters.

[0073] The process of constructing a historical processing dataset for the same scenario is as follows: set unified retrieval parameters, which include pipe material, cross-section type, processing technology type, cutting head model and CNC system type, and use the combination data of the unified retrieval parameters as a unique identifier for a processing scenario;

[0074] Based on the selection of unified search parameters, all matching data are extracted from the historical processing database. The matching data meets the requirements of having the same unified search parameters, no manual intervention in the processing process, and complete data records. The complete data records include cutting head posture, motion trajectory, collision situation, and control parameter information.

[0075] Integrate and match data to form a historical processing dataset of the same scenario, select a unified retrieval parameter as the key value, select matching data as the value, and integrate and store the data in the form of key-value pairs to obtain a set of historical CNC processing data for pipes.

[0076] In practice, the core configuration parameters of real-time pipe CNC machining are used as the retrieval basis. The stored entries in the historical processing database are matched one by one. Valid data with completely consistent parameters, no manual intervention in the processing process, and complete data records in all dimensions are strictly screened. The dataset of the same scene is standardized and integrated by relying on the key-value pair storage mode. The uniqueness judgment standard of the scene is strictly adhered to throughout the process to prevent the mixing of non-same source data across materials and processes. The foundation of scene adaptation analysis is solidified from the data source end.

[0077] The axial continuous motion trajectory of the cutting head in the historical CNC machining data set of pipes is segmented, and the feature data of each trajectory segment is extracted. The axial continuous motion trajectory of the cutting head represents the complete motion path of the cutting head along the pipe processing direction; the trajectory segment represents a continuous trajectory fragment divided according to the rotation angle threshold; the feature data includes the maximum lateral coverage of the cutting head, the length of the trajectory segment, the curvature of the trajectory segment, and the distance the trajectory segment is drawn when a collision occurs.

[0078] The segmentation process of the axial continuous motion trajectory of the cutting head is to check the angle value of each sampling point on the trajectory point by point, mark the sampling point whose angle value exceeds the angle threshold as the segment point, and divide the trajectory into multiple continuous small segments according to the interval of adjacent segment points.

[0079] The calculation of the maximum lateral coverage of the cutting head is based on the vertical plane of the direction of movement of the processing surface, which is perpendicular to the axis of pipe processing.

[0080] The length of a trajectory segment is calculated by obtaining the three-dimensional spatial coordinates of the starting and ending points of the trajectory segment, and then calculating the distance between the two points based on the spatial straight-line distance formula.

[0081] The curvature of a small segment of the trajectory is calculated by collecting the three-dimensional spatial coordinates of all sampling points on the small segment of the trajectory, and then calculating the curvature value point by point based on the curvature calculation formula and taking the arithmetic mean.

[0082] The distance drawn by the trajectory segment during the collision is extracted to locate the three-dimensional spatial coordinates of the collision start point and the collision end point in the trajectory segment, and the trajectory arc length between the two points is calculated.

[0083] The extraction process of all feature data is based on the small trajectory segments that have collided in the historical CNC machining data set of pipe materials. The extraction results are used as training samples for the association mapping model.

[0084] In practice, the characteristics of the continuous running trajectory of the cutting head along the pipe axis are matched, and the angle change range of the trajectory sampling points is verified point by point. The segmentation point is marked by the angle threshold that is suitable for the current scene, and the trajectory segment is refined. At the same time, the core features of the trajectory segment are calculated one by one by combining the three-dimensional spatial coordinate calculation rules. All feature extraction is carried out only around the historical collision trajectory, and invalid data interference from non-collision trajectory is filtered out, so that the extracted feature samples are completely consistent with the collision trajectory running rules in this scene.

[0085] Based on the analysis of the control parameters and trajectory segment features in historical data, a two-way correlation mapping relationship is obtained, forming a correlation mapping model between parameters and trajectory segment features. The control parameters include the cutting head attitude angle and motion speed.

[0086] The process of segmenting the continuous axial motion trajectory of the cutting head and extracting feature data to form an association mapping model includes:

[0087] A turning angle threshold is set as the segmentation criterion. When the turning angle of any point on the continuous axial movement trajectory of the cutting head exceeds the turning angle threshold, that point is the segmentation point. The complete trajectory is divided into multiple continuous trajectory segments based on the segmentation point.

[0088] For each trajectory segment, the maximum lateral coverage of the cutting head is obtained by calculating the angle between the vector from the starting point to the ending point of the trajectory segment and the set running direction vector. The length of the trajectory segment is obtained by calculating the straight-line distance between the starting point and the ending point. The curvature of the trajectory segment is obtained by calculating the mean curvature of all points on the trajectory segment. For trajectory segments that collide, the distance drawn by the trajectory segment at the time of collision is obtained by extracting the trajectory length from the collision starting point to the collision ending point. Based on the two selection results, the correlation mapping model is analyzed through cross-validation.

[0089] The first screening result is to screen historical data segments with the same values ​​across all dimensions of the control parameters, compare and analyze the corresponding trajectory segment feature data, and statistically obtain the overlap interval of various trajectory segment feature data when the control parameters are fixed.

[0090] The second screening result is to screen historical data segments with the same short-segment feature data, extract the corresponding control parameters, and statistically obtain the value distribution of the control parameters when the trajectory is fixed.

[0091] Using a single trajectory segment as the smallest unit of analysis, the first screening result and the second screening result are cross-validated to establish a bidirectional correlation mapping relationship between the combination of control parameters and the feature data of the trajectory segment, and integrate them to form a correlation mapping model between parameters and the features of the trajectory segment.

[0092] In practical implementation, a single trajectory segment is taken as the smallest unit of analysis. The historical data association logic is broken down from two dimensions: constant control parameters and constant trajectory characteristics. The inherent linkage between control parameters and trajectory characteristics is sorted out through two-way cross-validation. The one-sidedness of single-dimensional modeling is abandoned, and a complete association mapping model with the current processing scenario is built. This ensures that the parameter association logic inside the model is fully adapted to the processing characteristics of the corresponding pipe material, and avoids the analysis bias caused by the generalization of the general model.

[0093] Based on the analysis of historical CNC machining data of pipes, the time required to predict the future trajectory of the cutting head is obtained, which is denoted as the prediction time. The motion data and control parameters of the cutting head that are currently running are collected and input into the trained machine learning model to predict the motion data of the cutting head within the prediction time.

[0094] Extract the running time span of all complete trajectories from the historical CNC machining data set of pipe materials. The running time span is the time difference between the start and end times of trajectory acquisition.

[0095] The extracted running time span is cleaned to remove outliers and retain the valid time span data. The arithmetic mean of the valid time span data is calculated and the result is used as the predicted duration of the future trajectory of the cutting head.

[0096] The training process of the machine learning model involves using the cutting head running data and control parameters of the collision trajectory that have occurred in the historical CNC machining data set of pipes as input samples, and the cutting head motion data within the corresponding time period as output samples to complete the training and verification of the model.

[0097] The motion data and control parameters of the cutting head during its current operating phase are collected. The motion data includes the three-dimensional spatial coordinates and attitude angles at each sampling time, and the control parameters include attitude angles and motion speed. These are input into a trained machine learning model, which outputs the cutting head motion data for the predicted duration.

[0098] In practice, we first sort out the time span data of the complete historical trajectories in the same scene, remove abnormal discrete values ​​and determine a reasonable trajectory prediction duration. Then, we use the full-dimensional running data of historical collision trajectories as training samples to complete the iterative training and effectiveness verification of the machine learning model. In actual operation, we collect the current coordinates, attitude and control parameters of the cutting head in real time, and rely on mature models to infer the trajectory data of future time periods. By matching the running sequence of the scene, we ensure the accuracy of the prediction results.

[0099] Input the cutting head motion data within the predicted time period into the correlation mapping model, extract the feature information corresponding to the predicted motion data for analysis, and make predictions and supervision of the risk of processing collisions.

[0100] Extract the feature information corresponding to the cutting head motion data within the prediction time. The extraction logic is the same as the extraction method of the feature data of the historical collision trajectory segment. The feature information includes the maximum lateral coverage, length and curvature of the predicted trajectory segment.

[0101] The extracted feature information is input into the parameter and mapped to the trajectory segment feature association model, and the feature data of all collision trajectories in the model are retrieved.

[0102] Calculate the deviation between the predicted feature information and the collision feature data in the model. The deviation is the absolute value of the numerical difference between the feature data of each dimension.

[0103] The deviation value is compared with a preset feature deviation threshold. When the deviation value in any dimension is less than the feature deviation threshold, it is determined that the predicted trajectory segment has a collision tendency.

[0104] After the collision tendency is determined, a predictive regulatory conclusion is generated, which includes the location of the trajectory segment with collision tendency, the deviation value of the corresponding dimension, and the matching collision feature data.

[0105] In practice, the feature extraction standard of historical collision trajectories is used to break down the core feature parameters of the predicted trajectory. These parameters are then compared with the scene-specific collision features built into the model. The potential collision tendency is determined by the feature deviation magnitude, high-risk trajectory segments are accurately located, and the corresponding deviation dimensions and matching features are analyzed to form a prediction conclusion. Quantitative comparison logic is used throughout the process to eliminate subjective judgment errors and achieve risk identification in the non-collision stage.

[0106] Based on the results of predictive monitoring, dynamic correction parameters are derived and transmitted to the CNC machining system to automatically adjust the cutting head's operating status. At the same time, real-time machining data is fed back and stored in the historical pipe CNC machining data set.

[0107] Based on the control parameters corresponding to the collision feature data that match the collision tendency trajectory in the correlation mapping model, and combined with the deviation value in the predictive regulatory conclusion, the dynamic correction parameters are derived. The dynamic correction parameters include the cutting head attitude angle adjustment value and the motion speed adjustment value.

[0108] The dynamic correction parameters are transmitted to the control module of the CNC machining system. After receiving the parameters, the control module automatically adjusts the attitude angle and movement speed of the cutting head, and the adjustment direction is to increase the deviation value between the predicted feature and the collision feature.

[0109] The real-time processing data feedback storage process involves classifying the currently processed trajectory data, control parameters, dynamic correction parameters, and predictive regulatory conclusions according to unified retrieval parameters;

[0110] The categorized real-time processing data is added to the historical CNC machining data set for pipes, and the retrieval index information of the data set is updated after the addition is completed;

[0111] In practice, the historical control parameters corresponding to the collision tendency are used as a benchmark. The adaptive dynamic correction parameters are derived by combining the characteristic deviation amplitude. The parameters are directly transmitted to the CNC machining system to complete the adaptive adjustment of the cutting head posture and running speed. The adjustment direction always revolves around widening the characteristic difference between the predicted trajectory and the collision trajectory. At the same time, the data of the entire machining process are classified and archived according to the scenario parameters, and the historical database retrieval index is updated synchronously to form a closed-loop iterative optimization mechanism to continuously improve the scenario adaptability and accuracy of collision avoidance monitoring.

[0112] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the invention. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, it is intended that all variations falling within the meaning and scope of equivalents of the claims be included within the present invention.

Claims

1. An intelligent anti-collision monitoring method applied to CNC machining of pipes, characterized in that: The intelligent collision avoidance monitoring method includes: Acquire historical processing data that matches the real-time processing scenario, and construct a historical processing dataset with the same scenario. The historical processing dataset with the same scenario is a set of historical pipe CNC processing data that conforms to the consistency of the processing scenario after being filtered according to unified search parameters. The axial continuous motion trajectory of the cutting head in the historical CNC machining data set of pipes is segmented, and the feature data of each trajectory segment is extracted. The axial continuous motion trajectory of the cutting head represents the complete motion path of the cutting head along the pipe processing direction; the trajectory segment represents a continuous trajectory fragment divided according to the rotation angle threshold; the feature data includes the maximum lateral coverage of the cutting head, the length of the trajectory segment, the curvature of the trajectory segment, and the distance the trajectory segment is drawn when a collision occurs. Based on the analysis of the control parameters and trajectory segment features in historical data, a two-way correlation mapping relationship is obtained, forming a correlation mapping model between parameters and trajectory segment features. The control parameters include the cutting head attitude angle and motion speed. Based on the analysis of historical CNC machining data of pipes, the time required to predict the future trajectory of the cutting head is obtained, which is denoted as the prediction time. The motion data and control parameters of the cutting head that are currently running are collected and input into the trained machine learning model to predict the motion data of the cutting head within the prediction time. Input the cutting head motion data within the predicted time period into the correlation mapping model, extract the feature information corresponding to the predicted motion data for analysis, and make predictions and supervision of the risk of processing collisions. The aforementioned risk assessment and monitoring of processing collisions includes: Extract the feature information corresponding to the cutting head motion data within the prediction time. The extraction logic is the same as the extraction method of the feature data of the historical collision trajectory segment. The feature information includes the maximum lateral coverage, length and curvature of the predicted trajectory segment. The extracted feature information is input into the parameter and mapped to the trajectory segment feature association model, and the feature data of all collision trajectories in the model are retrieved. Calculate the deviation between the predicted feature information and the collision feature data in the model. The deviation is the absolute value of the numerical difference between the feature data of each dimension. The deviation value is compared with a preset feature deviation threshold. When the deviation value in any dimension is less than the feature deviation threshold, it is determined that the predicted trajectory segment has a collision tendency. After the collision tendency is determined, a predictive regulatory conclusion is generated. The predictive regulatory conclusion includes the location of the trajectory segment with a collision tendency, the deviation value of the corresponding dimension, and the matching collision feature data. Based on the results of predictive monitoring, dynamic correction parameters are derived and transmitted to the CNC machining system to automatically adjust the cutting head's operating status. Simultaneously, real-time machining data is fed back and stored in the historical CNC machining data set for pipes. The dynamic correction parameters derived from the results of predictive supervision include: Based on the control parameters corresponding to the collision feature data that match the collision tendency trajectory in the correlation mapping model, and combined with the deviation value in the predictive regulatory conclusion, the dynamic correction parameters are derived. The dynamic correction parameters include the cutting head attitude angle adjustment value and the motion speed adjustment value. The dynamic correction parameters are transmitted to the control module of the CNC machining system. After receiving the parameters, the control module automatically adjusts the attitude angle and movement speed of the cutting head, and the adjustment direction is to increase the deviation value between the predicted feature and the collision feature. The real-time processing data feedback storage process involves classifying the currently processed trajectory data, control parameters, dynamic correction parameters, and predictive regulatory conclusions according to unified retrieval parameters; The categorized real-time processing data is added to the historical CNC machining data set for pipes. After the addition is completed, the retrieval index information of the data set is updated.

2. The intelligent anti-collision monitoring method for CNC machining of pipes according to claim 1, characterized in that: The association mapping model between the formation parameters and the features of the trajectory segments includes: The process of segmenting the continuous axial motion trajectory of the cutting head and extracting feature data to form an association mapping model includes: A turning angle threshold is set as the segmentation criterion. When the turning angle of any point on the continuous axial movement trajectory of the cutting head exceeds the turning angle threshold, that point is the segmentation point. The complete trajectory is divided into multiple continuous trajectory segments based on the segmentation point. For each trajectory segment, the maximum lateral coverage of the cutting head is obtained by calculating the angle between the vector from the starting point to the ending point of the trajectory segment and the set running direction vector. The length of the trajectory segment is obtained by calculating the straight-line distance between the starting point and the ending point. The curvature of the trajectory segment is obtained by calculating the mean curvature of all points on the trajectory segment. For trajectory segments that collide, the distance drawn by the trajectory segment at the time of collision is obtained by extracting the trajectory length from the collision starting point to the collision ending point. Based on the two selection results, the correlation mapping model is analyzed through cross-validation. The first screening result is to screen historical data segments with the same values ​​across all dimensions of the control parameters, compare and analyze the corresponding trajectory segment feature data, and statistically obtain the overlap interval of various trajectory segment feature data when the control parameters are fixed. The second screening result is to screen historical data segments with the same short-segment feature data, extract the corresponding control parameters, and statistically obtain the value distribution of the control parameters when the trajectory is fixed. Using a single trajectory segment as the smallest unit of analysis, the first and second screening results are cross-validated to establish a bidirectional correlation mapping relationship between the combination of control parameters and the feature data of the trajectory segment, and to integrate them into a correlation mapping model between parameters and the features of the trajectory segment.

3. The intelligent anti-collision monitoring method for CNC machining of pipes according to claim 1, characterized in that: The historical CNC machining data set for pipes that conforms to the consistency of the processing scenario includes: The process of constructing a historical processing dataset for the same scenario is as follows: set unified retrieval parameters, which include pipe material, cross-section type, processing technology type, cutting head model and CNC system type, and use the combination data of the unified retrieval parameters as a unique identifier for a processing scenario; Based on the selection of unified search parameters, all matching data are extracted from the historical processing database. The matching data meets the requirements of having the same unified search parameters, no manual intervention in the processing process, and complete data records. The complete data records include cutting head posture, motion trajectory, collision situation, and control parameter information. Integrate and match data to form a historical processing dataset for the same scenario. Select a unified retrieval parameter as the key and match data as the value. Integrate and store the data in the form of key-value pairs to obtain a set of historical CNC processing data for pipes.

4. The intelligent anti-collision monitoring method for CNC machining of pipes according to claim 1, characterized in that: The process involves segmenting the continuous axial motion trajectory of the cutting head in the historical CNC machining data set of pipes, breaking it down, and extracting the feature data of each trajectory segment, including: The segmentation process of the axial continuous motion trajectory of the cutting head is to check the angle value of each sampling point on the trajectory point by point, mark the sampling point whose angle value exceeds the angle threshold as the segment point, and divide the trajectory into multiple continuous small segments according to the interval of adjacent segment points. The calculation of the maximum lateral coverage of the cutting head is based on the vertical plane of the direction of movement of the processing surface, which is perpendicular to the axis of pipe processing. The length of a trajectory segment is calculated by obtaining the three-dimensional spatial coordinates of the starting and ending points of the trajectory segment, and then calculating the distance between the two points based on the spatial straight-line distance formula. The curvature of a small segment of the trajectory is calculated by collecting the three-dimensional spatial coordinates of all sampling points on the small segment of the trajectory, and then calculating the curvature value point by point based on the curvature calculation formula and taking the arithmetic mean. The distance drawn by the trajectory segment during the collision is extracted to locate the three-dimensional spatial coordinates of the collision start point and the collision end point in the trajectory segment, and the trajectory arc length between the two points is calculated. The extraction of all feature data is based on the trajectory segments that have collided in the historical CNC machining data set of pipes, and the extraction results are used as training samples for the association mapping model.

5. The intelligent anti-collision monitoring method for CNC machining of pipes according to claim 1, characterized in that: The prediction obtains the cutting head motion data within the predicted time period, including: Extract the running time span of all complete trajectories from the historical CNC machining data set of pipe materials. The running time span is the time difference between the start and end times of trajectory acquisition. The extracted running time span is cleaned to remove outliers and retain the valid time span data. The arithmetic mean of the valid time span data is calculated and the result is used as the predicted duration of the future trajectory of the cutting head. The training process of the machine learning model involves using the cutting head running data and control parameters of the collision trajectory that have occurred in the historical CNC machining data set of pipes as input samples, and the cutting head motion data within the corresponding time period as output samples to complete the training and verification of the model. The motion data and control parameters of the cutting head during its current operating phase are collected. The motion data includes the three-dimensional spatial coordinates and attitude angles at each sampling time, and the control parameters include attitude angles and motion speed. These are input into a trained machine learning model, which outputs the cutting head motion data for the predicted duration.

6. An intelligent anti-collision monitoring system for CNC machining of pipes, wherein the system is applied to the intelligent anti-collision monitoring method for CNC machining of pipes as described in any one of claims 1-5, characterized in that: The intelligent collision avoidance monitoring system includes a data set construction module, a trajectory feature processing module, a motion data prediction module, and a collision prediction and correction module. The data set construction module is used to acquire historical processing data that matches the real-time processing scenario and construct a historical processing dataset that conforms to the consistency of the processing scenario. The trajectory feature processing module is used to segment the continuous axial motion trajectory of the cutting head in the historical pipe CNC processing data set and extract the feature data of each trajectory segment. The motion data prediction module is used to determine the prediction duration based on the historical pipe CNC processing data set and predict the cutting head motion data within the prediction duration through a machine learning model. The collision prediction and correction module is used to predict the processing collision tendency based on the correlation mapping model, derive dynamic correction parameters, and feed back and store the real-time processing data.

7. The intelligent anti-collision monitoring system for CNC machining of pipes according to claim 6, characterized in that: The data set construction module includes a retrieval parameter setting unit and a matching data integration unit. The retrieval parameter setting unit is used to set unified retrieval parameters and use the combined data of the unified retrieval parameters as a unique identifier for a type of processing scenario. The matching data integration unit is used to extract matching data based on the unified retrieval parameters and integrate it into a historical pipe CNC processing data set in the form of key-value pairs. The trajectory feature processing module includes a trajectory segmentation execution unit and a feature data extraction unit. The trajectory segmentation execution unit is used to verify the turning angle value of each sampling point on the trajectory point by point, and to mark the sampling points that exceed the turning angle threshold as segmentation points, thus dividing the trajectory into continuous small segments. The feature data extraction unit is used to calculate and extract the maximum horizontal coverage of the cutting head, the length of the small trajectory segment, the curvature of the small trajectory segment, and the distance the small trajectory segment is drawn when a collision occurs.

8. The intelligent anti-collision monitoring system for CNC machining of pipes according to claim 6, characterized in that: The motion data prediction module includes a prediction duration calculation unit and a motion data output unit. The prediction duration calculation unit is used to extract the running time span of historical trajectories, and calculates and determines the prediction duration of the future trajectory of the cutting head after data cleaning. The motion data output unit is used to train a machine learning model, collect the motion data and control parameters of the cutting head that have been running, input them into the model, and output the motion data of the cutting head within the prediction duration. The collision prediction and correction module includes a collision tendency determination unit and a correction parameter feedback unit. The collision tendency determination unit is used to extract feature information from the predicted motion data, calculate the deviation value from the collision feature data, and compare it with a preset threshold to determine the collision tendency of the trajectory segment. The correction parameter feedback unit is used to derive dynamic correction parameters based on the corresponding control parameters in the associated mapping model and the deviation value, transmit them to the CNC machining system, and feed back and store real-time machining data.

Images