Thermal extension error compensation method and system for turning and milling head

Abstract

The present invention discloses a method and system for compensating for thermal stretch errors of a milling head, and relates to the technical field of machining. The method comprises: obtaining a reference size and a size tolerance bandwidth of a target workpiece, and retrieving a workpiece sequence for size measurement and image acquisition. Analyze and obtain size trend information, including a size deviation value and a trend mark. Configure a verification recognition window according to a first trend step, retrieve a workpiece surface image to obtain a verification workpiece surface image. Perform bidirectional differential recognition on the verification workpiece surface image to obtain a verification compensation recognition window. Determine whether the difference between the verification compensation recognition window and the first compensation recognition window meets a preset value. If so, use the size deviation value and the trend mark as targets to perform error compensation; otherwise, adjust the first trend step until the difference between the recognition window meets the preset value. This solves the technical problems of error compensation lag and poor compensation effect, and achieves the technical effects of improving error compensation responsiveness and improving machining accuracy.

Description

Technical Field

[0001] The invention relates to the technical field of machining, and in particular to a method and system for compensating thermal extension errors of a milling head. Background Art

[0002] The turning and milling swing head is an important component used to adjust the angle between the machine tool spindle and the workpiece surface during the machining process. In the actual machining process, due to the temperature rise during turning and milling, the swing head component will be affected by the machining heat source and produce thermal deformation. When the turning and milling swing head is thermally deformed, the machining accuracy will be reduced, affecting the machining quality.

[0003] At present, the degree of thermal deformation of the milling head is mainly identified by the processing data of the processed parts. However, this method has large hysteresis and error, and cannot obtain accurate error compensation. There are technical problems such as error compensation lag and poor compensation effect. Summary of the invention

[0004] The present invention provides a method and system for compensating thermal extension errors of a turning and milling head, so as to solve the technical problems of error compensation lag and poor compensation effect in the prior art, and achieve the technical effects of improving error compensation responsiveness and improving machining accuracy.

[0005] In a first aspect, the present invention provides a method for compensating a thermal extension error of a milling head, wherein the method comprises:

[0006] Interactively obtain the benchmark size information and size tolerance bandwidth of the target workpiece;

[0007] Retrieving a target processing workpiece sequence within a preset processing monitoring period, and using a dimension measurement module and a surface detection module to sequentially perform dimension measurement and image acquisition on the target processing workpiece sequence to obtain a dimension measurement data sequence and a workpiece surface image sequence, wherein the target processing workpiece sequence has a processing time mark;

[0008] Performing a dimension trend analysis on the dimension measurement data sequence, the reference dimension information, and the dimension tolerance bandwidth according to a first trend step length to obtain a first dimension deviation value and a first dimension trend mark, wherein the first dimension trend mark includes an increment mark and a decrement mark, and the first dimension deviation value corresponds to a first compensation identification window;

[0009] Configure a first verification recognition window based on the first compensation recognition window, and retrieve the workpiece surface image sequence with the first verification recognition window as an index to obtain a verification workpiece surface image sequence;

[0010] Using a surface defect identifier to perform bidirectional differential recognition on the verification workpiece surface image sequence to obtain a verification compensation recognition window;

[0011] Determine whether the recognition window difference between the verification compensation recognition window and the first compensation recognition window meets a preset window difference, and if so, use the first size deviation value and the first size trend mark as the target size deviation value and the target size trend mark;

[0012] If not, adjusting the first trend step size according to the identification window difference until the identification window difference meets the preset window difference;

[0013] Error compensation is performed based on the target size deviation value and the target size trend indicator.

[0014] In this method, by combining the analysis of dimensional measurement data and surface images, a precise path for obtaining workpiece dimensions and surface quality is provided. Through dimensional trend analysis and identification window difference judgment, real-time monitoring and accuracy judgment of workpiece dimension deviation are achieved. When the deviation data is accurate data, error compensation is performed, thereby improving the processing accuracy and quality stability of the processed workpiece and reducing processing errors and defective product rates.

[0015] In a second aspect, the present invention further provides a milling head thermal extension error compensation system, wherein the system comprises:

[0016] A dimension information acquisition module, wherein the dimension information acquisition module is used to interactively obtain reference dimension information and dimension tolerance bandwidth of a target workpiece;

[0017] A sample sequence acquisition module, the sample sequence acquisition module is used to retrieve a target processing workpiece sequence within a preset processing monitoring period, and use a dimension measurement module and a surface detection module to sequentially perform dimension measurement and image acquisition on the target processing workpiece sequence to obtain a dimension measurement data sequence and a workpiece surface image sequence, wherein the target processing workpiece sequence has a processing time mark;

[0018] a trend analysis module, the trend analysis module being used to perform a dimension trend analysis on the dimension measurement data sequence, the reference dimension information and the dimension tolerance bandwidth according to a first trend step length, to obtain a first dimension deviation value and a first dimension trend mark, wherein the first dimension trend mark includes an increment mark and a decrement mark, and the first dimension deviation value corresponds to a first compensation identification window;

[0019] A verification retrieval module, the verification retrieval module is used to configure a first verification recognition window based on the first compensation recognition window, and to retrieve the workpiece surface image sequence with the first verification recognition window as an index to obtain a verification workpiece surface image sequence;

[0020] A verification compensation recognition module, the verification compensation recognition module is used to perform bidirectional differential recognition on the verification workpiece surface image sequence using a surface defect identifier to obtain a verification compensation recognition window;

[0021] A comparison and determination module, wherein the comparison and determination module is used to determine whether the identification window difference between the verification compensation identification window and the first compensation identification window satisfies a preset window difference, and if so, the first size deviation value and the first size trend mark are used as the target size deviation value and the target size trend mark;

[0022] A step length adjustment module, wherein if no, the step length adjustment module is used to adjust the first trend step length according to the identification window difference until the identification window difference meets the preset window difference;

[0023] An error compensation module is used to perform error compensation based on the target size deviation value and the target size trend indicator.

[0024] The present invention discloses a method and system for compensating thermal extension errors of a milling head, including: interactively obtaining reference dimension information and dimension tolerance bandwidth of a target processed workpiece; calling a target processed workpiece sequence within a preset processing monitoring period, and using a dimension measurement module and a surface detection module to sequentially perform dimension measurement and image acquisition on the target processed workpiece sequence to obtain a dimension measurement data sequence and a workpiece surface image sequence, wherein the target processed workpiece sequence has a processing time mark; performing dimension trend analysis on the dimension measurement data sequence, the reference dimension information and the dimension tolerance bandwidth according to a first trend step length to obtain a first dimension deviation value and a first dimension trend mark, wherein the first dimension trend mark includes an increment mark and a decrement mark, and the first dimension deviation value corresponds to a first compensation mark. Identification window; configure the first verification identification window based on the first compensation identification window, use the first verification identification window as an index, retrieve the workpiece surface image sequence, and obtain the verification workpiece surface image sequence; use the surface defect identifier to perform bidirectional differential identification on the verification workpiece surface image sequence to obtain the verification compensation identification window; determine whether the identification window difference between the verification compensation identification window and the first compensation identification window meets the preset window difference, if so, use the first size deviation value and the first size trend mark as the target size deviation value and the target size trend mark; if not, adjust the first trend step according to the identification window difference until the identification window difference meets the preset window difference; perform error compensation based on the target size deviation value and the target size trend mark. The method and system for compensating for thermal extension errors of the turning and milling swing head disclosed in the present invention solve the technical problems of error compensation lag and poor compensation effect, and achieve the technical effect of improving error compensation responsiveness and improving machining accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 It is a schematic flow chart of the method for compensating the thermal extension error of the turning and milling head of the present invention;

[0026] Figure 2 It is a structural schematic diagram of the thermal extension error compensation system of the turning and milling swing head of the present invention.

[0027] Explanation of the reference numerals: size information acquisition module 11, sample sequence acquisition module 12, trend analysis module 13, verification retrieval module 14, verification compensation identification module 15, comparison and discrimination module 16, step adjustment module 17, error compensation module 18. DETAILED DESCRIPTION

[0028] The technical solution provided in the embodiments of the present invention is to solve the technical problems of error compensation lag and poor compensation effect in the prior art. The overall idea adopted is as follows:

[0029] First, interact with the user to obtain the reference size information and size tolerance bandwidth of the target workpiece. Then, retrieve the target workpiece sequence within the preset processing monitoring period. Then, use the size measurement module and the surface detection module to perform size measurement and image acquisition on the target workpiece sequence in turn to obtain a size measurement data sequence and a workpiece surface image sequence, wherein the target workpiece sequence has a processing time mark. Next, according to the set first trend step, the size measurement data sequence, the reference size information and the size tolerance bandwidth are analyzed for size trend, and the first size deviation value and the first size trend mark are obtained, wherein the first size trend mark includes an increment mark and a decrement mark, and the first size deviation value corresponds to the first compensation recognition window. Then, configure the first verification recognition window according to the first compensation recognition window, retrieve the workpiece surface image sequence with the first verification recognition window as the index, and obtain the verification workpiece surface image sequence. Then, use the surface defect identifier to perform bidirectional differential recognition on the verification workpiece surface image sequence to obtain the verification compensation recognition window. Finally, determine whether the recognition window difference between the verification compensation recognition window and the first compensation recognition window meets the preset window difference. If yes, the first size deviation value and the first size trend mark are used as the target size deviation value and the target size trend mark; if no, the first trend step is adjusted according to the recognition window difference until the recognition window difference meets the preset window difference. Error compensation is performed based on the target size deviation value and the target size trend mark.

[0030] The above technical solution will be described in detail below in conjunction with the accompanying drawings and specific implementation methods of the specification to better understand the above technical solution. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments of the present invention. It should be understood that the present invention is not limited to the example embodiments used only to explain the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention. In addition, it should be noted that for the convenience of description, only the parts related to the present invention are shown in the drawings, rather than all of them.

[0031] Embodiment 1

[0032] Figure 1 The figure is a flow chart of the method for compensating the thermal extension error of the turning and milling head of the present invention; wherein the method is applied to the error compensation system, the error compensation system is communicatively connected with the dimension measurement module and the surface detection module, and the method comprises:

[0033] Interactively obtain the benchmark size information and size tolerance bandwidth of the target workpiece;

[0034] The target workpiece refers to the original workpiece to be milled, and the reference size information and size tolerance bandwidth of the target workpiece.

[0035] Optionally, the reference size information of the target workpiece is obtained, and the reference size information includes the size parameters of the target workpiece before processing and the processing target, such as length, width, height, etc. In other words, the reference size information reflects the processing target and processing starting point of the target workpiece.

[0036] Optionally, the dimensional tolerance bandwidth refers to the allowable range or tolerance of the size when manufacturing or designing a product. It indicates the allowable change or error range of the product size during the manufacturing process. Specifically, the dimensional tolerance bandwidth can be divided into positive tolerance and negative tolerance. Positive tolerance refers to the range in which the size is allowed to be slightly larger than the design size when manufacturing a product. Negative tolerance refers to the range in which the size is allowed to be slightly smaller than the design size. The size of the dimensional tolerance bandwidth depends on the product requirements and manufacturing process.

[0037] Through the above steps, it is ensured that the accurate benchmark dimension information and dimension tolerance bandwidth of the target workpiece are obtained, providing accurate reference data for subsequent processing.

[0038] Retrieving a target processing workpiece sequence within a preset processing monitoring period, and using a dimension measurement module and a surface detection module to sequentially perform dimension measurement and image acquisition on the target processing workpiece sequence to obtain a dimension measurement data sequence and a workpiece surface image sequence, wherein the target processing workpiece sequence has a processing time mark;

[0039] Optionally, the dimension measurement module is used to measure the dimension of the target workpiece, save the measurement data as a data sequence, and obtain a dimension measurement data sequence. Exemplarily, the dimension measurement module includes a laser rangefinder, a coordinate measuring machine and other equipment. The surface detection module includes an image acquisition device, an image processing device and an image storage device, and the surface detection module is used to acquire the surface image of the target workpiece. Exemplarily, the workpiece surface image is acquired by a camera, a microscope and other equipment, and the image is processed and analyzed, and finally the processed image is saved as a workpiece surface image sequence.

[0040] Optionally, the target processing workpiece sequence has a processing time mark, which can be an absolute time mark or a relative time mark, which is realized by associating and storing the processing time information with the dimension measurement data sequence and the workpiece surface image sequence in the data saving step. Among them, the absolute time mark refers to the time mark under the standard world time coordinate, including the specific date and time, and the relative time mark is the time interval relative to a certain starting time point, such as the time interval relative to the start time of processing, which can include units such as seconds, minutes, hours, etc. In addition, the choice of using an absolute time mark or a relative time mark depends on the specific situation and analysis requirements.

[0041] Optionally, the dimensional measurement data sequence includes dimensional measurement data of multiple milling cycles from the start of processing of the target workpiece to the current moment, wherein a milling cycle refers to the time range for the milling equipment to perform a complete milling operation on the target processing surface of the target workpiece. In other words, the milling cycle is the time window for the milling equipment to reduce material per unit thickness on the target workpiece. In a narrow sense, it is the time for milling a layer. After multiple continuous milling cycles, the target workpiece finally achieves the processing goal.

[0042] Performing a dimension trend analysis on the dimension measurement data sequence, the reference dimension information, and the dimension tolerance bandwidth according to a first trend step length to obtain a first dimension deviation value and a first dimension trend mark, wherein the first dimension trend mark includes an increment mark and a decrement mark, and the first dimension deviation value corresponds to a first compensation identification window;

[0043] Among them, the first dimension deviation value refers to the value that best represents the deviation of the processing dimension within the preset processing monitoring period, and this value is a scalar. The first dimension trend mark indicates the deviation direction corresponding to the deviation value. Specifically, it includes: the incremental mark, which means that the processing dimension is getting larger than the reference dimension, indicating that the milling amount is insufficient; the decrement mark, which means that the processing dimension is getting smaller than the reference dimension, indicating that the milling amount is too large.

[0044] In some embodiments, a dimension trend analysis is performed on the dimension measurement data sequence, the reference dimension information, and the dimension tolerance bandwidth according to a first trend step length to obtain a first dimension deviation value and a first dimension trend mark, wherein the first dimension trend mark includes an increment mark and a decrement mark, and the first dimension deviation value corresponds to a first compensation identification window. The method includes:

[0045] Traversing the dimension measurement data sequence and performing difference calculation with the reference dimension information to obtain a dimension deviation data sequence;

[0046] sequentially determine whether the size deviation data sequence is greater than or equal to the size deviation data of the size tolerance bandwidth, if so, generate an abnormal size deviation data sequence, if not, generate a tolerance size deviation data sequence;

[0047] Performing a size trend analysis on the abnormal size deviation data sequence according to the first trend step length to obtain an abnormal compensation identification window;

[0048] Performing a size trend analysis on the tolerance size deviation data sequence according to the first trend step length to obtain a tolerance compensation identification window;

[0049] The first compensation identification window is obtained by fusing the abnormal compensation identification window and the tolerant compensation identification window.

[0050] Optionally, to obtain a dimension deviation data sequence, first, based on the processing time mark and the feed amount sequence, the reference dimension information is cumulatively removed to obtain a standard dimension data sequence. The standard dimension data sequence includes ideal dimension data of multiple processing time marks corresponding to the dimension measurement data sequence, reflecting the dimension change of the target workpiece under ideal processing conditions. Then, the dimension measurement data sequence and the standard dimension data sequence are traversed to perform difference calculation to obtain a dimension deviation data sequence including dimension deviation data at multiple time marks.

[0051] Optionally, determine in turn whether the dimension deviation data sequence is greater than or equal to the dimension deviation data of the dimension tolerance bandwidth. If it is greater, it means that the error at the time mark in the dimension deviation data sequence exceeds the tolerance of the target processing workpiece, and the dimension deviation data is stored in the abnormal dimension deviation data sequence. Otherwise, it is stored in the tolerant dimension deviation data sequence.

[0052] Among them, the abnormal compensation identification window refers to the time period where the abnormal size is concentrated, and the first trend step is the time length of a single movement in space when performing size trend analysis, which is set by the technician or the expert system. In other words, the first trend step is the step length of the sliding window movement of the size trend analysis. Through trend analysis, the area where the abnormal size deviation is concentrated is found to further accurately identify the abnormal time period.

[0053] Optionally, based on the same acquisition method of the abnormal compensation identification window, a tolerance compensation identification window is obtained. The tolerance compensation identification window reflects the part with larger deviation in the size deviation data that meets the size tolerance bandwidth. The tolerance compensation identification window serves as an expansion window for deviation compensation, which helps to improve the reliability of deviation compensation.

[0054] Furthermore, by fusing the abnormal compensation identification window and the tolerance compensation identification window, a first compensation identification window can be obtained for further deviation compensation analysis.

[0055] In some embodiments, the method comprises:

[0056] Taking the first compensation identification window as an index, extracting the size deviation data sequence to obtain a first size deviation data sequence;

[0057] Performing size mean processing on the first size deviation data sequence to obtain the first size deviation value;

[0058] Determine whether the first dimension deviation value is greater than the reference dimension information, and if so, the first dimension tendency mark is an incremental mark;

[0059] If not, the first size tendency indicator is a reduction indicator.

[0060] Optionally, to determine whether the first dimension trend mark is an incremental mark or a decremental mark, first, use the first compensation identification window as an index to extract data from the dimension deviation data sequence to obtain a first dimension deviation data sequence. Then, perform mean processing on the first dimension deviation data sequence to obtain a first dimension deviation value. Then, determine whether the first dimension deviation value is greater than the reference dimension information. If so, the first dimension trend mark is an incremental mark. If not, the first dimension trend mark is a decremental mark. Through the above steps, the dimension deviation trend in the first compensation identification window is determined for subsequent deviation compensation processing.

[0061] In some implementations, performing a size trend analysis on the abnormal size deviation data sequence according to the first trend step length to obtain an abnormal compensation identification window, the method includes:

[0062] Constructing an abnormal two-dimensional space based on the abnormal size deviation data sequence, wherein the abnormal two-dimensional space has a plurality of abnormal coordinate points, and each abnormal coordinate point corresponds to an abnormal size deviation data;

[0063] Taking the center of the abnormal two-dimensional space as the starting point, moving according to the first trend step length to obtain a stage point, determining whether the density ratio of the stage point to the starting point is greater than 1, and if so, updating the stage point as the starting point and continuing to move;

[0064] When the density difference between the stage point and the starting point is less than the preset density difference, the movement is stopped, and the stage point corresponding to the maximum density during the movement is taken as the target abnormal coordinate point;

[0065] A time period determined by taking the time point corresponding to the target abnormal coordinate point as the center point and taking the first trend step length as the time interval radius is used as the abnormal compensation identification window.

[0066] Specifically, first, the abnormal size deviation data sequence is represented as a two-dimensional space. For example, a plane orthogonal coordinate system is constructed, in which each abnormal coordinate point in the plane orthogonal coordinate system corresponds to an abnormal size deviation data. Then, the center of the abnormal two-dimensional space, that is, the origin of the plane orthogonal coordinate system, is taken as the starting point, and the first trend step is used to move to obtain the stage point, and it is determined whether the density ratio of the stage point to the starting point is greater than 1. If so, the stage point is updated to the starting point and the movement continues. Among them, the first trend step is the positive direction of the time series.

[0067] Furthermore, until the density difference between the stage point and the starting point is less than the preset density difference, the movement is stopped, and the stage point corresponding to the maximum density during the movement is used as the target abnormal coordinate point, and the time point corresponding to the target abnormal coordinate point is used as the center point, and the time period determined by the first trend step as the time interval radius is used as the abnormal compensation identification window.

[0068] Configure a first verification recognition window based on the first compensation recognition window, and retrieve the workpiece surface image sequence with the first verification recognition window as an index to obtain a verification workpiece surface image sequence;

[0069] Optionally, configuring the first verification identification window based on the first compensation identification window includes extracting corresponding time nodes based on a plurality of milling cycles included in the first compensation identification window, and storing them as the first verification identification window.

[0070] Optionally, the first compensation recognition window is used as an index to retrieve the workpiece surface image sequence to obtain a verification workpiece surface image sequence, wherein multiple workpiece surface images in the verification workpiece surface image sequence correspond one-to-one to multiple milling cycles included in the first compensation recognition window.

[0071] Optionally, the image sequence of the verified workpiece surface is analyzed to identify the time period when the surface defects change. After the milling head undergoes thermal stretching, it will have a certain impact on the machining accuracy of the milling machine, which will lead to changes in surface roughness, flatness and other characteristics or the generation of burrs. Furthermore, the changes in the workpiece surface also change with the degree of thermal stretching. Therefore, through the changes in the workpiece surface, the time period when the surface defects change is identified, thereby verifying the first compensation recognition window and improving the accuracy of the dimensional deviation value.

[0072] Exemplarily, the identification of the time period of surface defect changes in the verification workpiece surface image sequence includes: preprocessing the image, such as denoising, graying, etc.; extracting the features of the surface defects, such as texture, color, shape, etc., from the preprocessed image; using feature matching algorithms (such as SIFT, SURF, etc.) to match the images in the image sequence to find similar image pairs; comparing the matched image pairs to detect changes in surface defects. Change detection can be performed using image difference comparison methods or machine learning methods; based on the results of change detection, the time period of surface defect changes is identified, and the identified time period of surface defect changes is used as the verification window for verifying the workpiece surface image sequence.

[0073] Using a surface defect identifier to perform bidirectional differential recognition on the verification workpiece surface image sequence to obtain a verification compensation recognition window;

[0074] In some embodiments, a surface defect identifier is used to perform bidirectional differential recognition on the verification workpiece surface image sequence to obtain a verification compensation recognition window, and the method includes:

[0075] The surface defect identifier includes a high-speed extraction branch, a low-speed extraction branch and a window extraction network layer;

[0076] matching a high-speed extraction interval and a low-speed extraction interval based on the first size deviation value, wherein the high-speed extraction interval is greater than the low-speed extraction interval;

[0077] Acquire multiple sample verification workpiece surface image sequences, high-speed extraction intervals, low-speed extraction intervals, multiple high-speed extraction image sets, multiple low-speed extraction image sets, multiple high-speed extraction defect feature sequences, multiple low-speed extraction feature sequences and multiple sample verification compensation recognition windows as a recognizer training data set;

[0078] Training the surface defect identifier using the identifier training data set;

[0079] The trained surface defect identifier is called to perform bidirectional differential recognition on the verification workpiece surface image sequence to obtain the verification compensation recognition window.

[0080] Optionally, a network structure including a high-speed extraction branch, a low-speed extraction branch and a window extraction network layer is used. The high-speed extraction branch is used to capture rapidly changing surface defect features, and the low-speed extraction branch is used to extract features of surface details. Among them, the high-speed extraction interval corresponding to the high-speed extraction branch is small, and the time interval between two adjacent images extracted is short. In other words, the sampling rate of the high-speed extraction branch is high, and it can capture rapidly changing feature information in a relatively large amount of data. Correspondingly, the low-speed extraction interval corresponding to the low-speed extraction branch is large, and the time interval between two adjacent images extracted is long. In other words, the sampling rate of the low-speed extraction branch is low, and a relatively small amount of data can be analyzed more finely.

[0081] Optionally, prepare a plurality of sample verification workpiece surface image sequences, and match the high-speed extraction interval and the low-speed extraction interval according to the known size deviation value. Generate a corresponding high-speed extraction image set and a low-speed extraction image set for each sample, and generate a corresponding high-speed extraction defect feature sequence and a low-speed extraction feature sequence. Use the prepared training data set to train the surface defect identifier so that it can accurately identify surface defects in the verification workpiece surface image sequence.

[0082] Optionally, a bidirectional differential recognition is performed on the surface image sequence of the verification workpiece based on the trained surface defect identifier to obtain a verification compensation recognition window. By analyzing the high-speed extraction feature sequence and the low-speed extraction feature sequence, the time period of surface defect change is identified to verify the accuracy of the first compensation recognition window.

[0083] In some implementations, the method includes:

[0084] Using the plurality of sample verification workpiece surface image sequences, high-speed extraction intervals and a plurality of high-speed extraction image sets as training data, to obtain the high-speed extraction branch;

[0085] Using the plurality of sample verification workpiece surface image sequences, low-speed extraction intervals and a plurality of low-speed extraction image sets as training data to obtain the low-speed extraction branch;

[0086] The high-speed extraction branch and the low-speed extraction branch are fully connected in parallel and connected to the input layer of the trained window extraction network layer to obtain the surface defect identifier.

[0087] Specifically, a surface defect identifier is constructed, and multiple sample verification workpiece surface image sequences, high-speed extraction intervals, and multiple high-speed extraction image sets are used as high-speed extraction branch training data, and multiple sample verification workpiece surface image sequences, low-speed extraction intervals, and multiple low-speed extraction image sets are used as low-speed extraction branch training data, and training is constructed separately to generate high-speed extraction branches and low-speed extraction branches. The high-speed extraction branch has a smaller network depth, thereby ensuring the response performance of the high-speed extraction, and the network depth of the low-speed extraction branch is relatively large, which helps to capture detailed features.

[0088] Furthermore, the above branches are fully connected in parallel and connected to the input layer of the window extraction network layer to form a complete surface defect identifier. This identifier will be able to use both fast and slow changing features to accurately identify defects on the surface of the verified workpiece and generate a verification compensation identification window.

[0089] In some implementations, the method includes:

[0090] Using the multiple high-speed extraction image sets and the multiple high-speed extraction defect feature sequences as training data, a high-speed feature extraction sub-network layer of the window extraction network layer is obtained;

[0091] Using the multiple low-speed extraction image sets and the multiple low-speed extraction defect feature sequences as training data, a low-speed feature extraction sub-network layer of the window extraction network layer is obtained;

[0092] Using the multiple high-speed defect feature extraction sequences, the multiple low-speed defect feature extraction sequences and the multiple sample verification compensation recognition windows as training data, to obtain a window recognition sub-network layer of the window extraction network layer;

[0093] The high-speed feature extraction sub-network layer is fully connected in parallel with the low-speed feature extraction sub-network layer, and then connected to the input layer of the window recognition sub-network layer to obtain the trained window extraction network layer.

[0094] The method in the above implementation uses the defect feature sequences of the high-speed extraction image set and the low-speed extraction image set as training data to train the high-speed feature extraction sub-network layer and the low-speed feature extraction sub-network layer of the window extraction network layer respectively. Then, these two sub-network layers are fully connected in parallel with the window recognition sub-network layer, and then connected to the input layer of the window recognition sub-network layer to construct a complete window extraction network layer. By considering the surface defect features at different time scales and training the high-speed and low-speed feature extraction sub-networks separately, the changes in surface defects can be better captured. The final window extraction network layer can effectively generate a verification compensation recognition window based on the identified defects on the workpiece surface.

[0095] Determine whether the recognition window difference between the verification compensation recognition window and the first compensation recognition window meets a preset window difference, and if so, use the first size deviation value and the first size trend mark as the target size deviation value and the target size trend mark;

[0096] Furthermore, if the verification compensation recognition window and the recognition window difference of the first compensation recognition window meet the preset window difference, the first size deviation value and the first size trend mark can be used as the target size deviation value and the target size trend mark, thereby ensuring that accurate size deviation values ​​and size trend marks are obtained, thereby improving the accuracy and reliability of compensation.

[0097] If not, adjusting the first trend step size according to the identification window difference until the identification window difference meets the preset window difference;

[0098] Optionally, if the identification window difference between the verification compensation identification window and the first compensation identification window does not meet the preset window difference, it means that the first trend step size is set unreasonably and needs to be adjusted. Furthermore, if the first trend step size is larger, more error details will be lost, and if the first trend step size is smaller, it cannot well reflect the overall deviation trend.

[0099] Error compensation is performed based on the target size deviation value and the target size trend indicator.

[0100] Optionally, error compensation is performed by adjusting processing parameters, correcting workpiece position, or adopting compensation technology. Exemplarily, based on a neural network model, with target size deviation value and target size trend identification as input data, and cutting parameter control scheme as output data, an error compensation model is constructed and trained. The neural network model forms error compensation capability by learning the mapping relationship between deviation conditions in sample data and control schemes.

[0101] In summary, the method for compensating the thermal extension error of the turning and milling head provided by the present invention has the following technical effects:

[0102] Obtain the reference dimension information and dimension tolerance bandwidth of the target workpiece through interaction; retrieve the target workpiece sequence within the preset processing monitoring period, and use the dimension measurement module and the surface detection module to perform dimension measurement and image acquisition on the target workpiece sequence in turn, to obtain a dimension measurement data sequence and a workpiece surface image sequence, wherein the target workpiece sequence has a processing time mark; perform dimension trend analysis on the dimension measurement data sequence, the reference dimension information and the dimension tolerance bandwidth according to the first trend step, to obtain a first dimension deviation value and a first dimension trend mark, wherein the first dimension trend mark includes an increment mark and a decrement mark, and the first dimension deviation value corresponds to a first compensation identification window; based on the first compensation identification window, the first compensation identification window is used to identify the first dimension deviation value and the first dimension trend mark. The window is configured with a first verification recognition window, and the first verification recognition window is used as an index to retrieve the workpiece surface image sequence to obtain the verification workpiece surface image sequence; the surface defect identifier is used to perform bidirectional differential recognition on the verification workpiece surface image sequence to obtain the verification compensation recognition window; it is determined whether the recognition window difference between the verification compensation recognition window and the first compensation recognition window meets the preset window difference, and if so, the first size deviation value and the first size trend mark are used as the target size deviation value and the target size trend mark; if not, the first trend step is adjusted according to the recognition window difference until the recognition window difference meets the preset window difference; error compensation is performed based on the target size deviation value and the target size trend mark. The technical effect of improving the responsiveness of error compensation and improving the processing accuracy is achieved.

[0103] Embodiment 2

[0104] Figure 2 Schematic diagram of the structure of the thermal extension error compensation system of the milling head of the present invention. Figure 1 The flow chart of the method for compensating the thermal extension error of the turning and milling head of the present invention can be shown as follows: Figure 2 The structure shown is implemented.

[0105] Based on the same concept as the method for compensating the thermal extension error of a turning and milling head in the above embodiment, the present invention also provides a system for compensating the thermal extension error of a turning and milling head, comprising:

[0106] A dimension information acquisition module 11 is used to interactively obtain reference dimension information and dimension tolerance bandwidth of a target workpiece;

[0107] The sample sequence acquisition module 12 is used to retrieve the target processing workpiece sequence within the preset processing monitoring period, and use the dimension measurement module and the surface detection module to perform dimension measurement and image acquisition on the target processing workpiece sequence in sequence to obtain a dimension measurement data sequence and a workpiece surface image sequence, wherein the target processing workpiece sequence has a processing time mark;

[0108] a trend analysis module 13, configured to perform a dimension trend analysis on the dimension measurement data sequence, the reference dimension information and the dimension tolerance bandwidth according to a first trend step length, to obtain a first dimension deviation value and a first dimension trend mark, wherein the first dimension trend mark includes an increment mark and a decrement mark, and the first dimension deviation value corresponds to a first compensation identification window;

[0109] A verification retrieval module 14 is used to configure a first verification recognition window based on the first compensation recognition window, and to retrieve the workpiece surface image sequence with the first verification recognition window as an index to obtain a verification workpiece surface image sequence;

[0110] A verification compensation identification module 15 is used to perform bidirectional differential identification on the verification workpiece surface image sequence using a surface defect identifier to obtain a verification compensation identification window;

[0111] A comparison and determination module 16 is used to determine whether the identification window difference between the verification compensation identification window and the first compensation identification window meets a preset window difference, and if so, use the first size deviation value and the first size trend mark as the target size deviation value and the target size trend mark;

[0112] a step length adjustment module 17, configured to adjust the first trend step length according to the identification window difference value, if not, until the identification window difference value satisfies a preset window difference value;

[0113] The error compensation module 18 is used to perform error compensation based on the target size deviation value and the target size trend indicator.

[0114] Wherein, the trend analysis module 13 includes:

[0115] A dimension deviation calculation unit, used for traversing the dimension measurement data sequence and performing difference calculation with the reference dimension information to obtain a dimension deviation data sequence;

[0116] A tolerance judgment unit is used to judge in turn whether the size deviation data sequence is greater than or equal to the size deviation data of the size tolerance bandwidth, and if so, generate an abnormal size deviation data sequence, and if not, generate a tolerant size deviation data sequence;

[0117] an abnormality analysis unit, configured to perform a size trend analysis on the abnormal size deviation data sequence according to the first trend step length to obtain an abnormality compensation identification window;

[0118] a tolerance analysis unit, configured to perform a size trend analysis on the tolerance size deviation data sequence according to the first trend step length to obtain a tolerance compensation identification window;

[0119] The window fusion unit is used to fuse the abnormal compensation recognition window and the tolerance compensation recognition window to obtain the first compensation recognition window.

[0120] Furthermore, the trend analysis module 13 also includes:

[0121] an index extraction unit, configured to extract the size deviation data sequence by taking the first compensation identification window as an index to obtain a first size deviation data sequence;

[0122] a size mean processing unit, configured to perform size mean processing on the first size deviation data sequence to obtain the first size deviation value;

[0123] an incremental identification unit, used to determine whether the first dimension deviation value is greater than the reference dimension information, and if so, the first dimension tendency identification is an incremental identification;

[0124] A decrement identification unit is used to identify the first size tendency as a decrement identification if no.

[0125] In some implementations, the anomaly analysis unit includes:

[0126] A coordinate point unit, used for constructing an abnormal two-dimensional space based on the abnormal size deviation data sequence, wherein the abnormal two-dimensional space has a plurality of abnormal coordinate points, and each abnormal coordinate point corresponds to an abnormal size deviation data;

[0127] a stage point moving unit, configured to take the center of the abnormal two-dimensional space as a starting point, move according to the first trend step length, obtain a stage point, determine whether the density ratio of the stage point to the starting point is greater than 1, and if so, update the stage point as the starting point and continue to move;

[0128] A target abnormal coordinate point setting unit is used to stop moving until the density difference between the stage point and the starting point is less than a preset density difference, and use the stage point corresponding to the maximum density during the movement as the target abnormal coordinate point;

[0129] an abnormal compensation identification window generating unit, used to take the time point corresponding to the target abnormal coordinate point as the center point and the time period determined by the first trend step length as the time interval radius as the abnormal compensation identification window

[0130] Furthermore, the verification compensation identification module 15 includes:

[0131] an extraction interval matching unit, configured to match a high-speed extraction interval and a low-speed extraction interval based on the first size deviation value, wherein the high-speed extraction interval is larger than the low-speed extraction interval;

[0132] A training data construction unit is used to obtain multiple sample verification workpiece surface image sequences, high-speed extraction intervals, low-speed extraction intervals, multiple high-speed extraction image sets, multiple low-speed extraction image sets, multiple high-speed extraction defect feature sequences, multiple low-speed extraction feature sequences and multiple sample verification compensation recognition windows as a recognizer training data set;

[0133] A defect identifier training unit, used to train the surface defect identifier using the identifier training data set;

[0134] The bidirectional differential recognition unit is used to call the trained surface defect identifier to perform bidirectional differential recognition on the verification workpiece surface image sequence to obtain the verification compensation recognition window.

[0135] Furthermore, the defect identifier training unit in the verification compensation identification module 15 further includes:

[0136] A high-speed extraction branch training unit is used to use the plurality of sample verification workpiece surface image sequences, high-speed extraction intervals and a plurality of high-speed extraction image sets as training data to obtain the high-speed extraction branch;

[0137] A low-speed extraction branch training unit is used to use the plurality of sample verification workpiece surface image sequences, the low-speed extraction intervals and the plurality of low-speed extraction image sets as training data to obtain the low-speed extraction branch;

[0138] The connection unit is used to fully connect the high-speed extraction branch and the low-speed extraction branch in parallel, and connect them with the input layer of the trained window extraction network layer to obtain the surface defect identifier.

[0139] Furthermore, the connection unit in the defect identifier training unit also includes:

[0140] A high-speed feature extraction sub-network unit, used to obtain a high-speed feature extraction sub-network layer of the window extraction network layer by using the multiple high-speed extraction image sets and the multiple high-speed extraction defect feature sequences as training data;

[0141] A low-speed feature extraction sub-network unit is used to obtain a low-speed feature extraction sub-network layer of the window extraction network layer by using the multiple low-speed extraction image sets and the multiple low-speed extraction defect feature sequences as training data;

[0142] A window recognition sub-network unit is used to use the multiple high-speed extraction defect feature sequences, the multiple low-speed extraction defect feature sequences and the multiple sample verification compensation recognition windows as training data to obtain a window recognition sub-network layer of the window extraction network layer;

[0143] The parallel fully connected unit is used to fully connect the high-speed feature extraction sub-network layer with the low-speed feature extraction sub-network layer in parallel, and then connect them with the input layer of the window recognition sub-network layer to obtain the trained window extraction network layer.

[0144] It should be understood that the embodiments mentioned in this specification focus on their differences from other embodiments. The specific embodiments in the aforementioned embodiment one are also applicable to the turning and milling head thermal expansion error compensation system described in embodiment two. For the sake of brevity of the specification, they will not be further elaborated here.

[0145] It should be understood that the embodiments disclosed in the present invention and the above description can enable those skilled in the art to use the present invention to implement the present invention. At the same time, the present invention is not limited to the above-mentioned embodiments. It should be understood that those skilled in the art can still modify the technical solutions recorded in the above-mentioned embodiments, or replace some of the technical features therein by equivalents; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the protection scope of the present invention.

Claims

1. A method for compensating the thermal extension error of a milling head, characterized in that: The method is applied to an error compensation system, the error compensation system is communicatively connected with a dimension measurement module and a surface detection module, and the method comprises: Interactively obtain the benchmark size information and size tolerance bandwidth of the target workpiece; Retrieving a target processing workpiece sequence within a preset processing monitoring period, and using a dimension measurement module and a surface detection module to sequentially perform dimension measurement and image acquisition on the target processing workpiece sequence to obtain a dimension measurement data sequence and a workpiece surface image sequence, wherein the target processing workpiece sequence has a processing time mark; Performing a dimension trend analysis on the dimension measurement data sequence, the reference dimension information, and the dimension tolerance bandwidth according to a first trend step length to obtain a first dimension deviation value and a first dimension trend mark, wherein the first dimension trend mark includes an increment mark and a decrement mark, and the first dimension deviation value corresponds to a first compensation identification window; Configure a first verification recognition window based on the first compensation recognition window, and retrieve the workpiece surface image sequence with the first verification recognition window as an index to obtain a verification workpiece surface image sequence; Using a surface defect identifier to perform bidirectional differential recognition on the verification workpiece surface image sequence to obtain a verification compensation recognition window; Determine whether the recognition window difference between the verification compensation recognition window and the first compensation recognition window meets a preset window difference, and if so, use the first size deviation value and the first size trend mark as the target size deviation value and the target size trend mark; If not, adjusting the first trend step size according to the identification window difference until the identification window difference meets the preset window difference; Error compensation is performed based on the target size deviation value and the target size trend indicator.

2. The method according to claim 1, characterized in that Performing a dimension trend analysis on the dimension measurement data sequence, the reference dimension information, and the dimension tolerance bandwidth according to a first trend step length to obtain a first dimension deviation value and a first dimension trend mark, wherein the first dimension trend mark includes an increment mark and a decrement mark, and the first dimension deviation value corresponds to a first compensation identification window, the method comprising: Traversing the dimension measurement data sequence and performing difference calculation with the reference dimension information to obtain a dimension deviation data sequence; sequentially determine whether the size deviation data sequence is greater than or equal to the size deviation data of the size tolerance bandwidth, if so, generate an abnormal size deviation data sequence, if not, generate a tolerance size deviation data sequence; Performing a size trend analysis on the abnormal size deviation data sequence according to the first trend step length to obtain an abnormal compensation identification window; Performing a size trend analysis on the tolerance size deviation data sequence according to the first trend step length to obtain a tolerance compensation identification window; The first compensation identification window is obtained by fusing the abnormal compensation identification window and the tolerant compensation identification window.

3. The method according to claim 2, characterized in that The method comprises: Taking the first compensation identification window as an index, extracting the size deviation data sequence to obtain a first size deviation data sequence; Performing size mean processing on the first size deviation data sequence to obtain the first size deviation value; Determine whether the first dimension deviation value is greater than the reference dimension information, and if so, the first dimension tendency mark is an incremental mark; If not, the first size tendency indicator is a reduction indicator.

4. The method according to claim 2, characterized in that Performing a size trend analysis on the abnormal size deviation data sequence according to the first trend step length to obtain an abnormal compensation identification window, the method comprising: Constructing an abnormal two-dimensional space based on the abnormal size deviation data sequence, wherein the abnormal two-dimensional space has a plurality of abnormal coordinate points, and each abnormal coordinate point corresponds to an abnormal size deviation data; Taking the center of the abnormal two-dimensional space as the starting point, moving according to the first trend step length to obtain a stage point, determining whether the density ratio of the stage point to the starting point is greater than 1, and if so, updating the stage point as the starting point and continuing to move; When the density difference between the stage point and the starting point is less than the preset density difference, the movement is stopped, and the stage point corresponding to the maximum density during the movement is taken as the target abnormal coordinate point; A time period determined by taking the time point corresponding to the target abnormal coordinate point as the center point and taking the first trend step length as the time interval radius is used as the abnormal compensation identification window.

5. The method according to claim 1, characterized in that Using a surface defect identifier to perform bidirectional differential recognition on the verification workpiece surface image sequence to obtain a verification compensation recognition window, the method includes: The surface defect identifier includes a high-speed extraction branch, a low-speed extraction branch and a window extraction network layer; matching a high-speed extraction interval and a low-speed extraction interval based on the first size deviation value, wherein the high-speed extraction interval is greater than the low-speed extraction interval; Acquire multiple sample verification workpiece surface image sequences, high-speed extraction intervals, low-speed extraction intervals, multiple high-speed extraction image sets, multiple low-speed extraction image sets, multiple high-speed extraction defect feature sequences, multiple low-speed extraction feature sequences and multiple sample verification compensation recognition windows as a recognizer training data set; Training the surface defect identifier using the identifier training data set; The trained surface defect identifier is called to perform bidirectional differential recognition on the verification workpiece surface image sequence to obtain the verification compensation recognition window.

6. The method according to claim 5, characterized in that The method comprises: Using the plurality of sample verification workpiece surface image sequences, high-speed extraction intervals and a plurality of high-speed extraction image sets as training data, to obtain the high-speed extraction branch; Using the plurality of sample verification workpiece surface image sequences, low-speed extraction intervals and a plurality of low-speed extraction image sets as training data to obtain the low-speed extraction branch; The high-speed extraction branch and the low-speed extraction branch are fully connected in parallel and connected to the input layer of the trained window extraction network layer to obtain the surface defect identifier.

7. The method according to claim 6, characterized in that The method comprises: Using the multiple high-speed extraction image sets and the multiple high-speed extraction defect feature sequences as training data, a high-speed feature extraction sub-network layer of the window extraction network layer is obtained; Using the multiple low-speed extraction image sets and the multiple low-speed extraction defect feature sequences as training data, a low-speed feature extraction sub-network layer of the window extraction network layer is obtained; Using the multiple high-speed defect feature extraction sequences, the multiple low-speed defect feature extraction sequences and the multiple sample verification compensation recognition windows as training data, to obtain a window recognition sub-network layer of the window extraction network layer; The high-speed feature extraction sub-network layer is fully connected in parallel with the low-speed feature extraction sub-network layer, and then connected to the input layer of the window recognition sub-network layer to obtain the trained window extraction network layer.

8. The thermal extension error compensation system of the milling head is characterized by: The system is used to implement the method described in any one of claims 1 to 7, the error compensation system is communicatively connected with the dimension measurement module and the surface detection module, and the system comprises: A dimension information acquisition module, wherein the dimension information acquisition module is used to interactively obtain reference dimension information and dimension tolerance bandwidth of a target workpiece; A sample sequence acquisition module, the sample sequence acquisition module is used to retrieve a target processing workpiece sequence within a preset processing monitoring period, and use a dimension measurement module and a surface detection module to sequentially perform dimension measurement and image acquisition on the target processing workpiece sequence to obtain a dimension measurement data sequence and a workpiece surface image sequence, wherein the target processing workpiece sequence has a processing time mark; a trend analysis module, the trend analysis module being used to perform a dimension trend analysis on the dimension measurement data sequence, the reference dimension information and the dimension tolerance bandwidth according to a first trend step length, to obtain a first dimension deviation value and a first dimension trend mark, wherein the first dimension trend mark includes an increment mark and a decrement mark, and the first dimension deviation value corresponds to a first compensation identification window; A verification retrieval module, the verification retrieval module is used to configure a first verification recognition window based on the first compensation recognition window, and to retrieve the workpiece surface image sequence with the first verification recognition window as an index to obtain a verification workpiece surface image sequence; A verification compensation recognition module, the verification compensation recognition module is used to perform bidirectional differential recognition on the verification workpiece surface image sequence using a surface defect identifier to obtain a verification compensation recognition window; A comparison and determination module, wherein the comparison and determination module is used to determine whether the identification window difference between the verification compensation identification window and the first compensation identification window satisfies a preset window difference, and if so, the first size deviation value and the first size trend mark are used as the target size deviation value and the target size trend mark; A step length adjustment module, wherein if no, the step length adjustment module is used to adjust the first trend step length according to the identification window difference until the identification window difference meets the preset window difference; An error compensation module is used to perform error compensation based on the target size deviation value and the target size trend indicator.

Images