Numerical control machine tool machining track real-time correction method based on multi-source visual perception

Abstract

The invention discloses a numerical control machine tool machining track real-time correction method based on multi-source visual perception, and the method comprises the following steps: S1, synchronously collecting a multi-source real-time image sequence of a machined workpiece in a machining process through a plurality of visual sensors disposed at different visual angles of a machining region of a numerical control machine tool; s2, performing preprocessing and feature extraction on the multi-source real-time image sequence, and fusing to obtain real-time three-dimensional coordinate information of key feature points, corresponding to the theoretical processing track model, on a processing workpiece; s3, the real-time three-dimensional coordinates of the key feature points are compared with theoretical coordinates of corresponding points in a pre-stored numerical control machine tool theoretical machining track model, and the track deviation between the actual machining track and the theoretical machining track is obtained through calculation; and S4, based on the trajectory deviation, generating a trajectory correction value through a preset correction algorithm. The method can better correct the machining track of the numerical control machine tool in real time.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of trajectory correction, in particular to a numerical control machine tool machining trajectory real-time correction method based on multi-source visual perception. BACKGROUND

[0002] In the machining process of numerical control machine tools, the precision of the machining trajectory directly determines the dimensional accuracy, shape accuracy and surface quality of the workpiece. However, in actual machining, due to factors such as mechanical transmission error, tool wear, load fluctuation, thermal deformation and external environmental interference, the actual machining trajectory is prone to deviate from the preset theoretical trajectory. If the deviation cannot be corrected in time, it may lead to workpiece scrap, reduced machining efficiency, and even equipment damage, so real-time correction of the machining trajectory is a key technical requirement to ensure the machining quality of numerical control machine tools.

[0003] In the existing machining trajectory correction method of numerical control machine tools, there are the following technical challenges: single perception means, relying mainly on a single sensor, which is difficult to obtain comprehensive two-dimensional details and three-dimensional spatial information of the workpiece, resulting in insufficient accuracy of feature point coordinate extraction; lack of unified coordinate reference for multi-source sensor data, and the conversion relationship between different sensor coordinate systems and machine tool coordinate systems is not clear, which may cause data fusion deviation and affect the accuracy of trajectory deviation calculation; single feedback control is often used in correction algorithms, which has a response lag problem and is difficult to adapt to the real-time compensation requirements of high-speed dynamic machining scenarios; there is a lack of effective monitoring of abnormal states during machining, and in the case of vibration, equipment abnormalities, etc. Blind correction may lead to error expansion, therefore, a numerical control machine tool machining trajectory real-time correction method based on multi-source visual perception is proposed. SUMMARY

[0004] The present application solves the above technical problems by the following technical solutions, the present application includes the following steps:

[0005] S1: Through multiple visual sensors deployed at different angles in the machining area of the numerical control machine tool, synchronously collect multi-source real-time image sequences of the machining workpiece during the machining process;

[0006] S2: Preprocess and extract features from the multi-source real-time image sequences, and fuse to obtain real-time three-dimensional coordinate information of key feature points on the machining workpiece corresponding to the theoretical machining trajectory model;

[0007] S3: Compare the real-time three-dimensional coordinates of the key feature points with the theoretical coordinates of the corresponding points in the pre-stored numerical control machine tool theoretical machining trajectory model, and calculate the trajectory deviation between the actual machining trajectory and the theoretical machining trajectory;

[0008] S4: Based on the trajectory deviation, generate a trajectory correction amount through a pre-set correction algorithm;

[0009] S5: feeding the trajectory correction amount to the control system of the numerical control machine tool in real time to dynamically compensate and correct the machining trajectory being executed.

[0010] Further, in step S1, the plurality of visual sensors include at least one high-resolution industrial camera and at least one laser displacement sensor; wherein the high-resolution industrial camera is used to acquire two-dimensional image information of the workpiece, and the laser displacement sensor is used to acquire three-dimensional point cloud data of the workpiece.

[0011] Further, in step S2, the specific acquisition process of real-time three-dimensional coordinate information includes:

[0012] S21: denoising and enhancing the two-dimensional image collected by the high-resolution industrial camera, and extracting the pixel coordinates of the key feature points corresponding to the theoretical machining trajectory model and the edges of the workpiece;

[0013] S22: registering and filtering the three-dimensional point cloud data collected by the laser displacement sensor, constructing a three-dimensional surface model of the workpiece, and extracting the three-dimensional space coordinates of the key feature points corresponding to the theoretical machining trajectory model;

[0014] S23: based on the camera calibration parameters obtained by offline calibration, converting the pixel coordinates of the key feature points extracted in step S21 into three-dimensional coordinates in the laser displacement sensor coordinate system, and data fusion with the three-dimensional space coordinates of the corresponding key feature points extracted in step S22 to obtain high-precision three-dimensional coordinates of the key feature points of the workpiece in the world coordinate system of the numerical control machine tool.

[0015] Further, in step S3, the specific process of calculating the trajectory deviation is:

[0016] The high-precision three-dimensional coordinates of the key feature points of the workpiece in the world coordinate system of the numerical control machine tool obtained by fusion are denoted as ;

[0017] The theoretical coordinates of the corresponding key feature points at the same time are obtained from the theoretical machining trajectory model, denoted as ;

[0018] The trajectory deviation vector of the key feature point is calculated as , and the calculation formula is:

[0019] ;

[0020] The trajectory deviation vector The components in each coordinate axis are the trajectory deviations.

[0021] Further in step S4, the preset correction algorithm is a correction algorithm based on a feedforward-feedback compound control; and a specific process of generating the trajectory correction amount is as follows: the trajectory deviation vector of the current time is input into a PID controller to calculate a feedback correction amount . The calculation formula is as follows:

[0022]

[0023] Wherein, , and are preset proportional, integral and differential coefficients, respectively.

[0024] Meanwhile, the theoretical motion speed and acceleration information of the current machining point are extracted according to the theoretical machining trajectory model to perform feedforward compensation and generate a feedforward correction amount

[0025] Finally, the feedforward correction amount and the feedback correction amount are added to obtain a total trajectory correction amount , and the specific process is as follows:

[0026] Further in step S5, the trajectory correction amount is fed back to the control system of the numerical control machine tool in real time, which is realized through an external mechanical origin offset function of the machine tool or a real-time Ethernet communication interface; and the trajectory correction amount is converted into a position compensation instruction or a speed compensation instruction of each motion axis of the numerical control machine tool, and is sent to the control system to complete dynamic compensation and correction of the machining trajectory.

[0027] Further in step S3, the process further comprises an offline calibration and modeling step:

[0028] Before machining starts, the multiple vision sensors are jointly calibrated using the same calibration object to determine the transformation relationship between the sensor coordinate system and the numerical control machine tool world coordinate system; and the theoretical machining trajectory model containing the theoretical coordinates of the key feature points is generated based on the CAD model of the workpiece.

[0029] Further in step S3, the process further comprises an offline calibration and modeling step: After the trajectory deviation vector of a single key feature point is calculated, the trajectory deviation vectors of multiple key feature points on the current machining section are statistically analyzed to calculate the average deviation vector and the variance; if the variance is greater than a preset threshold, it is determined that the machining process appears to be shaking or abnormal, and a warning signal is triggered and the correction process is paused.

[0030] ​​​Compared with the prior art, the present application has the following advantages: the multi-source visual perception-based numerical control machine tool machining track real-time correction method, by deploying a multi-source visual sensor composed of a high-resolution industrial camera and a laser displacement sensor, combining image denoising enhancement, point cloud registration filtering and coordinate fusion technology, obtaining high-precision three-dimensional coordinates of key feature points in a unified numerical control machine tool world coordinate system, providing accurate data basis for track deviation calculation; at the same time, based on the quantitative deviation calculation method of the actual and theoretical coordinate difference value, the accuracy of the deviation evaluation is ensured, and a reliable basis for correction is provided, a feedforward-feedback compound control algorithm is adopted, real-time feedback correction is realized through a PID controller, and feedforward compensation is combined with theoretical motion parameters to reduce correction lag; through the machine tool external mechanical origin offset function or real-time Ethernet communication interface, the correction amount is converted into motion shaft compensation instructions in real time, realizing dynamic real-time compensation of the machining track, quickly responding to the track deviation in the machining process, determining the transformation relationship between the sensor coordinate system and the machine tool world coordinate system through offline calibration, and generating a theoretical machining track model based on the CAD model, providing a unified and accurate reference for real-time correction, reducing real-time calculation error, and ensuring the consistency of the correction logic, through statistical analysis of the deviation of multiple key feature points in the machining section, the machining tremor or abnormality can be identified in time, triggering an early warning and pausing the correction, avoiding invalid correction or error expansion in abnormal state, and ensuring the safety of the machining process and the stability of the product quality, through the multi-source perception fusion, accurate deviation quantization, compound control correction and abnormal early warning technologies, the dynamic correction accuracy, real-time response speed and machining process reliability of the numerical control machine tool machining track are comprehensively improved, the machining error is effectively reduced, the product quality is ensured, and the system is more worthy of promotion and use. BRIEF DESCRIPTION OF DRAWINGS

[0031] Figure 1 is a flowchart of the present application. DETAILED DESCRIPTION

[0032] The embodiments of the present application will be described in detail below. The embodiments are implemented on the premise of the technical solutions of the present application, and detailed implementation methods and specific operation processes are given, but the protection scope of the present application is not limited to the following embodiments.

[0033] As shown in Figure 1 , the present embodiment provides a technical solution: a multi-source visual perception-based numerical control machine tool machining track real-time correction method, comprising the following steps:

[0034] S1: synchronously collecting multi-source real-time image sequences of the workpiece in the machining process by deploying multiple visual sensors at different angles in the machining area of the numerical control machine tool;

[0035] S2: preprocessing and feature extraction are performed on the multi-source real-time image sequence, and real-time three-dimensional coordinate information of key feature points corresponding to the theoretical machining trajectory model on the processed workpiece is obtained by fusion;

[0036] S3: the real-time three-dimensional coordinates of the key feature points are compared with the theoretical coordinates of the corresponding points in the pre-stored theoretical machining trajectory model of the numerical control machine tool, and the trajectory deviation between the actual machining trajectory and the theoretical machining trajectory is calculated;

[0037] S4: based on the trajectory deviation, a trajectory correction amount is generated through a pre-set correction algorithm;

[0038] S5: the trajectory correction amount is fed back to the control system of the numerical control machine tool in real time, and the dynamic compensation and correction of the machining trajectory being executed are performed.

[0039] In step S1, the plurality of vision sensors includes at least one high-resolution industrial camera and at least one laser displacement sensor; wherein the high-resolution industrial camera is used to acquire two-dimensional image information of the workpiece, and the laser displacement sensor is used to acquire three-dimensional point cloud data of the workpiece;

[0040] The high-resolution industrial camera is focused on acquiring two-dimensional image information of the workpiece, which can accurately capture the edge, surface texture and other planar detail features of the workpiece, providing clear image basis for pixel coordinate extraction of key feature points; the laser displacement sensor is specially used to acquire three-dimensional point cloud data of the workpiece, which can completely present the three-dimensional structure, depth information and surface topography of the workpiece, making up for the deficiency of two-dimensional image in spatial depth information. The combination of the two forms a multi-source data acquisition system, which comprehensively covers the two-dimensional details and three-dimensional structure information of the workpiece; the combination of the multi-source vision sensors provides high-quality raw data for subsequent preprocessing, feature extraction and coordinate fusion. The clear details of the two-dimensional image ensure the accuracy of the feature point pixel coordinate extraction, and the three-dimensional data of the three-dimensional point cloud support the reliability of the three-dimensional surface model construction and the spatial coordinate extraction, which together lay a data foundation for the acquisition of high-precision three-dimensional coordinates of key feature points, and indirectly ensure the accuracy of subsequent trajectory deviation calculation and correction.

[0041] The specific acquisition process of the real-time three-dimensional coordinate information in step S2 includes:

[0042] S21: denoising and enhancement processing are performed on the two-dimensional image collected by the high-resolution industrial camera, and the pixel coordinates of the key feature points corresponding to the theoretical machining trajectory model are extracted;

[0043] S22: the three-dimensional point cloud data collected by the laser displacement sensor is registered and filtered, a three-dimensional surface model of the workpiece is constructed, and the three-dimensional spatial coordinates of the key feature points corresponding to the theoretical machining trajectory model are extracted;

[0044] S23: Based on the camera calibration parameters obtained by offline calibration, convert the key feature point pixel coordinates extracted in step S21 into three-dimensional coordinates in the laser displacement sensor coordinate system, and perform data fusion with the corresponding key feature point three-dimensional space coordinates extracted in step S22 to obtain high-precision three-dimensional coordinates of the key feature points of the workpiece in the numerical control machine tool world coordinate system;

[0045] By denoising and enhancing the two-dimensional image of the high-resolution industrial camera, the image noise interference is effectively reduced, the details of the workpiece edge and key feature points are highlighted, and the accuracy of pixel coordinate extraction is ensured; the three-dimensional point cloud data of the laser displacement sensor is registered and filtered, which can remove redundant points and noise points, optimize the point cloud structure, and provide a high-quality data basis for three-dimensional surface model construction and key feature point three-dimensional space coordinate extraction. Based on the camera parameters calibrated offline, the pixel coordinates extracted from the two-dimensional image are converted to the laser displacement sensor coordinate system, and the corresponding feature point coordinates extracted from the three-dimensional point cloud are fused, and finally the high-precision three-dimensional coordinates in the numerical control machine tool world coordinate system are obtained. This process solves the coordinate system difference problem of multi-source sensor data, realizes the complementary advantages of two-dimensional detail information and three-dimensional space information, provides a unified and accurate coordinate reference for subsequent trajectory deviation calculation, ensures that the comparison between actual coordinates and theoretical coordinates has strict spatial consistency, and through the whole process optimization from preprocessing to feature extraction to coordinate fusion, the precision and stability of the three-dimensional coordinates of the key feature points are significantly improved. The reliable data input for subsequent trajectory deviation quantization calculation, correction generation and dynamic compensation makes the technical link from data acquisition to processing to correction form a closed loop, which directly guarantees the overall precision of real-time correction of the machining trajectory.

[0046] In step S3, the specific process of calculating the trajectory deviation is as follows:

[0047] The high-precision three-dimensional coordinates of the key feature points of the workpiece in the numerical control machine tool world coordinate system obtained by fusion are denoted as ;

[0048] The theoretical coordinates of the corresponding key feature points at the same time are obtained from the theoretical machining trajectory model, denoted as ;

[0049] The trajectory deviation vector of the key feature point is calculated as , and the calculation formula is:

[0050] ;

[0051] The trajectory deviation vector The components of the trajectory deviation vector in each coordinate axis are the trajectory deviation.

[0052] The trajectory deviation vector containing X, Y, Z axis components is obtained by directly subtracting the actual three-dimensional coordinates of the key feature points from the theoretical coordinates, which clearly quantifies the deviation degree of the actual machining trajectory from the theoretical trajectory in each dimension in space, solves the problem of ambiguous description of the deviation, makes the deviation size and direction directly measurable and accurately analyzable, and the components of the deviation vector in each coordinate axis directly reflect the error of the machining trajectory in different directions, providing accurate raw deviation data for the subsequent feed-forward-feedback composite correction algorithm. The PID controller can calculate the feedback correction amount in real time based on these quantized deviations, ensuring that the correction direction and amplitude match the actual deviation and avoiding blind correction. The consistency and objectivity of deviation evaluation are ensured: through a unified coordinate difference calculation formula, the deviation calculation standard is consistent for different machining stages and different key feature points, reducing human evaluation errors and providing reliable data support for the stability analysis and quality traceability of the machining process.

[0053] For example, in the scenario of milling a workpiece plane with a numerical control machine tool:

[0054] Based on the theoretical machining trajectory model, the theoretical coordinates of a key feature point (such as the workpiece surface point corresponding to the milling cutter cutting position) at a certain time are (unit: mm, corresponding to the X, Y, Z axes of the numerical control machine tool world coordinate system).

[0055] Through multi-source data fusion, the actual three-dimensional coordinates of the same key feature point at that time are obtained as .

[0056] According to the calculation method, the trajectory deviation vector ;

[0057] The deviation vector clearly shows that the actual trajectory deviates by 0.02 mm in the positive X-axis direction, 0.01 mm in the negative Y-axis direction, and 0.01 mm in the positive Z-axis direction. The subsequent correction algorithm can accurately calculate the compensation amount based on these three components, for example, applying a position compensation of -0.02 mm to the X-axis and +0.01 mm to the Y-axis, to finally achieve dynamic correction of the trajectory deviation.

[0058] In step S4, the preset correction algorithm is a feed-forward-feedback composite control-based correction algorithm; the specific process of generating the trajectory correction amount is: inputting the trajectory deviation vector at the current time into the PID controller to calculate the feedback correction amount , and the calculation formula is:

[0059] ;

[0060] wherein , and are the preset proportional, integral, and differential coefficients, respectively.

[0061] At the same time, the theoretical motion speed and acceleration information of the current processing point are extracted according to the theoretical processing trajectory model to perform feedforward compensation and generate the feedforward correction amount. ;

[0062] Finally, the feedforward correction amount is added to the feedback correction amount to obtain the total trajectory correction amount , specifically ;

[0063] Feedback correction uses the PID controller to calculate compensation in real time based on the current trajectory deviation vector, directly correcting existing deviations. Feedforward correction, on the other hand, generates compensation in advance based on the theoretical machining trajectory's motion parameters, such as velocity and acceleration, to anticipate deviations that may arise during machining due to inertia and delays. The combination of these two methods significantly reduces the lag inherent in pure feedback control and improves tracking accuracy for dynamic machining trajectories. Optimizing correction stability and interference rejection: The PID controller's proportional, integral, and differential components are used to quickly respond to deviations, eliminate accumulated errors, and suppress overshoot, respectively, ensuring feedback correction stability. Feedforward compensation, by incorporating theoretical motion information in advance, reduces the impact of external disturbances (such as load changes and tool wear) on the correction process, making the overall correction process more robust. Adapting to complex dynamic machining scenarios: In high-speed, variable-operation CNC machine tool machining, a single correction method struggles to achieve both speed and accuracy. The combined control algorithm, through the coordinated feedforward prediction and feedback correction, accurately adapts to rapidly changing trajectories and ensures consistent machining trajectories.

[0064] Continuous milling processing scenario: It is known that the trajectory deviation vector of the key feature point at a certain moment is mm (X, Y, Z axis deviation), the theoretical motion parameters of the key feature points at this moment in the theoretical processing trajectory model are: X-axis theoretical speed mm / s, acceleration mm / s²; Y-axis theoretical speed mm / s, acceleration mm / s²; Z-axis theoretical speed mm / s, acceleration mm / s²;

[0065] Feedback correction calculation: The preset PID controller parameters are , , (Assuming the current cumulative deviation , deviation change rate .

[0066] Feedback correction amount Calculated as: ;

[0067] ;

[0068] ;

[0069] i.e. .

[0070] Total correction amount:

[0071] .

[0072] The total correction amount is converted into position compensation instructions for each axis and then sent to the control system, which not only corrects the current deviation that has occurred, but also compensates for the subsequent deviation that may be caused by the inertia of movement in advance, so as to finally realize the precise dynamic correction of the machining trajectory.

[0073] In step S5, the real-time feedback of the trajectory correction amount to the control system of the numerical control machine tool is realized through the external mechanical origin offset function of the machine tool or the real-time Ethernet communication interface; the trajectory correction amount is converted into position compensation instructions or speed compensation instructions for each movement axis of the numerical control machine tool, and is sent to the control system to complete the dynamic compensation and correction of the machining trajectory;

[0074] The trajectory correction amount is fed back through an external mechanical origin offset function of the machine tool or a real-time Ethernet communication interface, both of which are industrial-level reliable transmission means. The real-time Ethernet communication interface can meet the data transmission requirements of high frequency and low delay, and the external mechanical origin offset function provides a hardware-level fast response path, both of which ensure that the correction amount can be transmitted to the control system in real time, avoid correction lag caused by transmission delay, and adapt to high-speed dynamic machining scenarios of the numerical control machine tool. The precise adaptation and execution of the correction instruction are realized: the total trajectory correction amount is converted into position compensation instructions or speed compensation instructions of each motion axis of the numerical control machine tool, so that the abstract correction amount is directly mapped to the executable motion control parameters of the machine tool. The position compensation instruction can accurately correct the instantaneous position deviation of the axis, and the speed compensation instruction can dynamically adjust the motion rate of the axis to offset the accumulated error. The two types of instructions adapt to the error characteristics of different machining stages as needed, ensure that the correction effect directly acts on the machining trajectory, and improve the compensation accuracy. The closed-loop integrity of the correction process is constructed, the generated correction amount is effectively converted into the actual action of the machine tool through a clear feedback mechanism and instruction conversion rule, the whole process from obtaining the deviation through multi-source perception to dynamic compensation correction forms a closed loop, which guarantees the operability and effectiveness of the whole real-time correction method, and avoids the technical fault caused by the fact that the correction amount cannot be implemented after being generated. The device compatibility and practicality of the method are enhanced: the external mechanical origin offset function and the real-time Ethernet communication interface are general technical interfaces in the numerical control machine tool industry, which can be adapted without large-scale modification of the machine tool, reduce the threshold of technology landing, make the real-time correction method compatible with numerical control machine tools of different brands and models, and improve the practical application value of the method.

[0075] Before step S3, an offline calibration and modeling step is further included:

[0076] Before the machining starts, the multiple vision sensors are jointly calibrated using the same calibration object to determine the transformation relationship between the sensor coordinate systems and the numerical control machine tool world coordinate system; and a theoretical machining trajectory model containing the theoretical coordinates of the key feature points is generated based on the CAD model of the workpiece;

[0077] The multiple vision sensors are jointly calibrated through the same calibration object to determine the transformation relationship between the sensor coordinate systems and the numerical control machine tool world coordinate system, which solves the problem of coordinate system consistency caused by installation position and angle difference of the multi-source vision sensors. This provides a strict spatial reference for the fusion of two-dimensional image pixel coordinates and three-dimensional point cloud coordinates, and the comparison of actual coordinates and theoretical coordinates in the subsequent steps, avoids error in deviation calculation caused by non-uniform coordinates, and guarantees the accuracy of data processing from the source.

[0078] The theoretical machining trajectory model is generated based on the CAD model of the workpiece, ensuring that the theoretical trajectory strictly adheres to the design requirements of the workpiece, providing accurate and traceable reference standards for deviation evaluation in real-time machining. The key feature point theoretical coordinates included in the theoretical model serve as direct comparison benchmarks for deviation calculation, providing clear basis for quantitative evaluation of the actual trajectory deviation from the theoretical trajectory, and avoiding correction direction deviation caused by ambiguous theoretical reference.

[0079] The calibration and modeling steps are completed offline before machining, without occupying the computing resources during real-time machining. After the time-consuming operations such as complex coordinate system transformation calculation and theoretical trajectory model generation are completed in advance, the real-time machining stage only needs to call the preset parameters and models, significantly reducing the real-time calculation pressure of the control system, ensuring that the real-time steps such as multi-source data acquisition, deviation calculation, and correction amount generation can run efficiently, adapting to the dynamic response requirements of high-speed machining of numerical control machine tools.

[0080] Strengthening the rigor and repeatability of the method: The standardized offline calibration process and theoretical modeling method based on CAD models make the reference setting of the entire correction method reproducible and verifiable. The correction process for different machining batches or different workpieces can be based on unified calibration parameters and modeling logic, reducing the fluctuation of machining errors caused by the randomness of reference setting, and improving the stability and reliability of the method in actual production.

[0081] After calculating the trajectory deviation vector of a single key feature point , the trajectory deviation vectors of multiple key feature points on the current machining segment are also statistically analyzed to calculate their average deviation vector and variance. If the variance is greater than the preset threshold, it is determined that the machining process has appeared to be shaking or abnormal, and a warning signal is triggered and the correction process is paused;

[0082] By calculating the average deviation and variance of the trajectory deviation vectors of multiple key feature points on the current machining segment, the variance index can effectively reflect the dispersion degree of the deviation. When the variance is greater than the preset threshold, it indicates that there may be problems such as shaking, tool abnormality, or load fluctuation in the machining process, which can timely convert implicit machining abnormalities into quantifiable warning signals, solving the problem of difficulty in identifying systematic abnormalities by relying solely on single-point deviation.

[0083] After determining the machining abnormality, a warning is triggered and the correction process is paused, which can prevent the correction algorithm from generating incorrect compensation instructions based on abnormal deviations in unstable states such as shaking and interference, and avoid further deviation of the trajectory or the generation of new machining errors caused by blind correction. This mechanism protects the machining quality and reduces the scrap rate.

[0084] Based on statistical analysis of multi-feature point deviation, compared with single feature point instantaneous deviation, the overall stability of the processing section can be better reflected, the risk of misjudgment caused by accidental error is reduced, and the abnormality judgment is more objective and reliable. At the same time, the early warning and suspension mechanism provides a time window for the operator to intervene in the investigation, which is convenient for timely troubleshooting of equipment failure and adjustment of process parameters, and ensures the controllability of the processing process.

[0085] In addition, the terms "first", "second", "third", etc. are used only for descriptive purposes and are not to be construed as indicating or implying relative importance or an indicated number of technical features. Therefore, the features defined with "first", "second", etc. can explicitly or implicitly include at least one of the features. In the description of the present application, the meaning of "a plurality of" is at least two, such as two, three, etc., unless otherwise specifically limited.

[0086] In the description of the present application, the description of the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In the present application, the illustrative description of the above terms is not necessarily directed to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any appropriate manner in any one or more embodiments or examples. In addition, different embodiments or examples described in the present application and the features of different embodiments or examples can be combined and combined by those skilled in the art without contradiction.

[0087] Although the embodiments of the present application have been shown and described above, it should be understood that the above embodiments are exemplary and should not be construed as limiting the present application, and those skilled in the art can make changes, modifications, replacements and variations to the above embodiments within the scope of the present application.

Claims

1. A method for real-time correction of machining trajectory of a numerical control machine tool based on multi-source visual perception, characterized in that, The method comprises the following steps: S1: synchronously collecting a plurality of real-time image sequences of a workpiece in a machining process through a plurality of visual sensors arranged at different angles in a machining area of a numerical control machine tool; S2: pre-processing and feature extraction are performed on the plurality of real-time image sequences, and real-time three-dimensional coordinate information of key feature points corresponding to a theoretical machining track model on the workpiece is obtained through fusion; S3: comparing the real-time three-dimensional coordinates of the key feature points with theoretical coordinates of corresponding points in the pre-stored theoretical machining track model of the numerical control machine tool, and calculating a track deviation between an actual machining track and the theoretical machining track; S4: generating a track correction amount through a preset correction algorithm based on the track deviation; S5: feeding back the track correction amount to a control system of the numerical control machine tool in real time, and dynamically compensating and correcting an executing machining track.

2. The method for real-time correction of machining trajectory of a CNC machine tool based on multi-source visual perception according to claim 1, characterized in that: In step S1, the plurality of visual sensors comprises at least one high-resolution industrial camera and at least one laser displacement sensor; wherein the high-resolution industrial camera is used to acquire two-dimensional image information of the workpiece, and the laser displacement sensor is used to acquire three-dimensional point cloud data of the workpiece.

3. The method of claim 2, wherein the method further comprises: The specific acquisition process of the real-time three-dimensional coordinate information in step S2 comprises: S21: performing denoising and enhancement processing on the two-dimensional image acquired by the high-resolution industrial camera, and extracting pixel coordinates of edges of the workpiece and key feature points corresponding to the theoretical machining track model; S22: performing registration and filtering processing on the three-dimensional point cloud data acquired by the laser displacement sensor, constructing a three-dimensional surface model of the workpiece, and extracting three-dimensional space coordinates of the key feature points corresponding to the theoretical machining track model; S23: based on camera calibration parameters obtained through offline calibration, converting the pixel coordinates of the key feature points extracted in step S21 into three-dimensional coordinates in a laser displacement sensor coordinate system, and performing data fusion with the three-dimensional space coordinates of the corresponding key feature points extracted in step S22, to obtain high-precision three-dimensional coordinates of the key feature points of the workpiece in a numerical control machine tool world coordinate system.

4. The method for real-time correction of machining trajectory of a CNC machine tool based on multi-source visual perception according to claim 3, characterized in that: In step S3, the specific process of calculating the track deviation is: The high-precision three-dimensional coordinates of the fused workpiece key feature points in the numerical control machine tool world coordinate system are marked as ; acquire the theoretical coordinates of the key feature points corresponding to the same moment from the theoretical machining trajectory model, denoted as ; calculating a trajectory deviation vector of the key feature point The formula is: ； Trajectory deviation vector The components in each coordinate axis are the trajectory deviations.

5. The method for real-time correction of machining trajectory of a CNC machine tool based on multi-source visual perception according to claim 4, characterized in that: In the step S4, the preset correction algorithm is a correction algorithm based on a feedforward-feedback compound control; and a specific process of generating the trajectory correction amount is as follows: inputting the trajectory deviation vector of the current moment to a PID controller to calculate a feedback correction amount , and the calculation formula is as follows: ； wherein, , and are preset proportional, integral and derivative coefficients, respectively. Meanwhile, the theoretical motion speed and acceleration information of the current machining point are extracted according to the theoretical machining track model to generate a feedforward correction amount ; Finally, the feedforward correction amount and the feedback correction amount are added to obtain a total trajectory correction amount , specifically .

6. The method for real-time correction of machining trajectory of a CNC machine tool based on multi-source visual perception according to claim 5, characterized in that: In step S5, the real-time feedback of the trajectory correction amount to the control system of the numerical control machine tool is achieved through an external mechanical origin offset function of the machine tool or a real-time Ethernet communication interface; the trajectory correction amount is converted into position compensation instructions or speed compensation instructions of each motion axis of the numerical control machine tool, and is sent to the control system to complete dynamic compensation and correction of the machining trajectory.

7. The method for real-time correction of machining trajectory of CNC machine tools based on multi-source visual perception according to claim 1, characterized in that: Before step S3, the method further comprises an offline calibration and modeling step: Before the machining starts, the plurality of visual sensors are jointly calibrated using the same calibration object, the transformation relationship between the sensor coordinate systems and the numerical control machine tool world coordinate system is determined, and the theoretical machining track model containing the theoretical coordinates of the key feature points is generated based on a CAD model of the workpiece.

8. The method for real-time correction of machining trajectory of a CNC machine tool based on multi-source visual perception according to claim 4, characterized in that: In calculating the trajectory deviation vector of the single key feature point Afterwards, the trajectory deviation vectors of multiple key feature points on the current machining section are statistically analyzed to calculate the average deviation vector and variance thereof. If the variance is greater than a preset threshold, it is determined that the machining process has a tremor or an abnormality, and a warning signal is triggered and the correction process is paused.

Images