A method, system and medium for predicting the position of a machine tool feed axis worktable
By dividing the training data sets of different screw states and polynomial interpolation methods, combined with the parameterized simulation model, the problems of low position prediction accuracy and poor efficiency in the case of screw motion state transition of traditional methods are solved, and high-precision machine tools are realized Predict the position of the shaft table.
Patent Information
- Application Number
- CN202411288849.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-14
- Publication Date
- 2024-12-31
- Estimated Expiration
- 2044-09-14
AI Technical Summary
In the case of screw motion state transition, a single machine learning model is difficult to accurately predict residual errors, resulting in low position accuracy and poor calculation efficiency.
By obtaining historical data, dividing it into training data sets of different screw states, training multiple machine learning models, and using polynomial interpolation method to plan imaginary trajectories, combining parameterized simulation models to predict the final position.
It improves the position prediction accuracy and calculation efficiency of the machine tool feed axis, and is suitable for high-precision repeat positioning under different motion states.
Smart Images
Figure CN118819066B_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the technical field of machine tool feed axis worktables, and in particular to a position prediction method, system and medium for a machine tool feed axis worktable. Background Art
[0002] The traditional method of predicting the position of the feed axis table of a machine tool is to obtain the basic predicted position of the feed axis table of the machine tool by solving the parameterized model based on the dynamic and kinematic models of the feed axis of the machine tool. There is a certain deviation between the basic predicted position solved by the parameterized model and the actual position of the feed axis table of the machine tool, that is, the residual error. Machine learning is then used to compensate for the residual error between the basic predicted position of the parameterized model and the actual position of the feed axis table of the machine tool.
[0003] However, when the machine tool feed axis screw is in motion and the command speed reaches V=0, the servo motor often experiences the following state transitions: (1) the motor reverses; (2) the motor maintains the rotation direction of the previous state.
[0004] When the above-mentioned state transition occurs, when a single machine learning model is used to predict the residual error, since the motion state of the feed axis screw is not classified and the transition of the motion state is not considered, it is easy for the model to try to find a way to distinguish these samples that cannot actually be distinguished by given features, causing the model to be confused and unable to describe the characteristics of the physical machine tool feed axis well, resulting in the residual error predicted by the machine learning model having problems such as low position accuracy and poor computational efficiency. Summary of the invention
[0005] In view of the above-mentioned deficiencies in the prior art, the present application provides a position prediction method, system and medium for a machine tool feed axis worktable, so as to solve the problem that the solution does not take into account the conversion of the motion state of the screw, and the residual error predicted by the machine learning model has the problems of low position accuracy and poor calculation efficiency.
[0006] In a first aspect, the present application provides a method for predicting the position of a feed axis table of a machine tool, the method comprising:
[0007] Obtaining command information issued to the servo system corresponding to the feed axis worktable of the machine tool in the historical time period, and inputting the command information into a preset parameterized simulation model to obtain an initial predicted position corresponding to each command information; wherein the command information includes a command position and a command speed;
[0008] Collect historical data corresponding to the feed axis worktable of the machine tool within a historical time period; wherein the historical data is composed of a plurality of data groups, each of which includes at least: instruction information, an initial predicted position and a residual error, and the residual error is the difference between the actual worktable position collected and the initial predicted position;
[0009] Taking the data group where the instruction speed is 0 in the historical data as the breakpoint, the historical data of the historical time period is divided into several data segments; subtracting the residual error of the first data group of the current data segment from the residual error in each data group in each data segment, and determining the subtracted data segment as the training data segment;
[0010] Based on the time sequence of the historical time periods, the order between the training data segments is determined; then the state conversion of the lead screw corresponding to the current training data segment and the previous training data segment is determined, and then several training data segments are divided into four training data sets; wherein the lead screw state is divided into a tension state and a compression state; the corresponding initial machine learning model is trained through the training data set to obtain four trained machine learning models;
[0011] Acquire a trajectory to be predicted; wherein the trajectory to be predicted is composed of instruction information to be predicted, and the instruction information to be predicted is composed of an instruction position to be predicted and an instruction speed to be predicted;
[0012] Inputting the command information to be predicted into a preset parameterized simulation model to obtain an initial predicted position of the trajectory to be predicted;
[0013] The predicted trajectory is divided into several predicted segments with the command speed of 0 as the breakpoint; the screw state conversion corresponding to each predicted segment of the predicted trajectory is obtained, and each predicted segment is input into the corresponding trained machine learning model to obtain the relative residual error of each predicted segment of the predicted trajectory; based on the tool setting point position, the first predicted command position of each predicted segment, and the maximum value of the predicted command speed in the current predicted segment, a preset hypothetical trajectory from the tool setting point position to the first predicted command position of each predicted segment is planned using the polynomial interpolation method; wherein the preset hypothetical trajectory is composed of hypothetical command information, and the hypothetical command information is composed of hypothetical command position and hypothetical command speed; the screw state conversion of the first hypothetical command position of the preset hypothetical trajectory corresponding to the same predicted segment and the screw state conversion of the first predicted command position of the predicted trajectory are the same;
[0014] Inputting the hypothetical command information of the preset hypothetical trajectory into the preset parameterized simulation model to obtain the initial predicted position of the preset hypothetical trajectory; taking the preset hypothetical trajectory as a breakpoint at a command speed of 0 to obtain a plurality of hypothetical trajectory segments;
[0015] Obtain the state conversion of the lead screw corresponding to each imaginary trajectory segment, and then input each imaginary trajectory segment into the corresponding trained machine learning model to obtain a preset error set corresponding to the preset imaginary trajectory; wherein the preset error set is composed of preset errors corresponding to each imaginary instruction information in the imaginary trajectory segment;
[0016] Based on the specific number of hypothetical trajectory segments, determining a residual error of a first command position of a current segment to be predicted;
[0017] The final predicted position is obtained based on the initial predicted position of the trajectory to be predicted, the relative residual errors of each segment and the residual errors of the first command position of each segment to be predicted.
[0018] Furthermore, based on the screw state changes corresponding to the current training data segment and the previous training data segment, several training data segments are divided into four training data sets, specifically including: based on the information instructions corresponding to the current training data segment and the previous training data segment, determining the screw state changes of the current training data segment; wherein the screw state changes are divided into tension-compression state, tension-tension state, compression-compression state, and compression-tension state; and then based on the four screw state changes, the training data segments are divided into corresponding training data sets.
[0019] Further, based on the initial predicted position of the trajectory to be predicted, the relative residual error of each segment and the residual error of the first command position of each segment to be predicted, the final predicted position is obtained, specifically including: by formula:
[0020] ,
[0021] The final predicted position H is calculated; where h represents the initial predicted position, j represents the jth trajectory segment to be predicted, and Z represents the total number of trajectory segments to be predicted. represents the relative residual error of each segment to be predicted of the jth predicted trajectory, represents the residual error of the first command position of the jth trajectory segment to be predicted.
[0022] In a second aspect, the present application provides a position prediction system for a machine tool feed axis worktable, the system comprising:
[0023] A historical data acquisition module is used to obtain the instruction information issued to the servo system corresponding to the machine tool feed axis worktable in the historical time period, and input the instruction information into a preset parameterized simulation model to obtain the initial predicted position corresponding to each instruction information; wherein the instruction information includes the instruction position and the instruction speed; collect the historical data corresponding to the machine tool feed axis worktable in the historical time period; wherein the historical data is composed of a plurality of data groups, each of which at least includes: instruction information, initial predicted position and residual error, and the residual error is the difference between the actual worktable position collected and the initial predicted position;
[0024] The training data acquisition module is used to divide the historical data of the historical time period into several data segments by taking the data group where the instruction speed is 0 in the historical data as the breakpoint; subtract the residual error of the first data group of the current data segment from the residual error in each data group in each data segment, and determine the subtracted data segment as the training data segment;
[0025] A data set division module is used to determine the order between training data segments based on the time sequence of historical time segments; further determine the screw state conversion status corresponding to the current training data segment and the previous training data segment, and further divide several training data segments into four training data sets; wherein the screw state is divided into a tension state and a compression state; train the corresponding initial machine learning model through the training data set to obtain four trained machine learning models;
[0026] A predicted position acquisition module is used to acquire a trajectory to be predicted; wherein the trajectory to be predicted is composed of instruction information to be predicted, and the instruction information to be predicted is composed of an instruction position to be predicted and an instruction speed to be predicted; the instruction information to be predicted is input into a preset parameterized simulation model to obtain an initial predicted position of the trajectory to be predicted;
[0027] The hypothetical data acquisition module is used to divide the issued to-be-predicted instruction information into several to-be-predicted segments with the instruction speed of 0 as the breakpoint of the to-be-predicted trajectory; obtain the state conversion of the lead screw corresponding to each to-be-predicted segment of the to-be-predicted trajectory, input each to-be-predicted segment into the corresponding trained machine learning model, and obtain the relative residual error of each to-be-predicted segment of the to-be-predicted trajectory; based on the tool setting point position, the first to-be-predicted instruction position of each to-be-predicted segment, and the maximum value of the to-be-predicted instruction speed in the current to-be-predicted segment, the polynomial interpolation method is used to plan the distance from the tool setting point position to each to-be-predicted segment. A preset hypothetical trajectory of a first command position to be predicted; wherein the preset hypothetical trajectory is composed of hypothetical command information, and the hypothetical command information is composed of hypothetical command position and hypothetical command speed; the state transition of the lead screw at the first hypothetical command position of the preset hypothetical trajectory corresponding to the same segment to be predicted is the same as the state transition of the lead screw at the first command position to be predicted of the trajectory to be predicted; the hypothetical command information of the preset hypothetical trajectory is input into a preset parameterized simulation model to obtain an initial predicted position of the preset hypothetical trajectory; the preset hypothetical trajectory is taken as a breakpoint at a command speed of 0 to obtain a plurality of hypothetical trajectory segments;
[0028] A residual error acquisition module is used to obtain the state conversion of the lead screw corresponding to each imaginary trajectory segment, and then input each imaginary trajectory segment into the corresponding trained machine learning model to obtain a preset error set corresponding to the preset imaginary trajectory; wherein the preset error set is composed of preset errors corresponding to each imaginary instruction information in the imaginary trajectory segment; based on the specific number of imaginary trajectory segments, determine the residual error of the first instruction position of the current segment to be predicted;
[0029] The final position acquisition module is used to obtain the final predicted position based on the initial predicted position of the trajectory to be predicted, the relative residual error of each segment and the residual error of the first command position of each segment to be predicted.
[0030] Further, the residual error acquisition module includes an error calculation unit, which is used to determine the preset error yn corresponding to the last hypothetical instruction position as the residual error of the first instruction position of the current segment to be predicted when the specific number of hypothetical trajectory segments is 1, the preset error set is composed of the preset errors of the only hypothetical trajectory segment: {y1, y2, y3...yn}; when the specific number of hypothetical trajectory segments m is greater than 1, the preset error set is composed of the preset errors of several hypothetical trajectory segments: {( , , … )、( , , … )、...( , , … )}, determine the sum of the preset errors corresponding to the last imaginary command position of each imaginary trajectory segment is the residual error of the first instruction position of the current segment to be predicted.
[0031] Furthermore, the data set division module includes a division unit, which is used to determine the screw state change of the current training data segment based on the information instructions corresponding to the current training data segment and the previous training data segment; wherein the screw state change is divided into tension-compression state, tension-tension state, compression-compression state, and compression-tension state; and then based on the four screw state changes, the training data segment is divided into corresponding training data sets.
[0032] Furthermore, the final position acquisition module includes a final position calculation unit,
[0033] Used by the formula:
[0034] ,
[0035] The final predicted position H is calculated; where h represents the initial predicted position, j represents the jth trajectory segment to be predicted, and Z represents the total number of trajectory segments to be predicted. represents the relative residual error of each segment to be predicted of the jth predicted trajectory, represents the residual error of the first command position of the jth trajectory segment to be predicted.
[0036] In a third aspect, the present application provides a non-volatile computer storage medium having computer instructions stored thereon, which, when executed, implement a method for predicting the position of a machine tool feed axis table as described above.
[0037] Those skilled in the art can understand that the present application has at least the following beneficial effects:
[0038] 1. Considering the different tension and compression characteristics of the feed shaft screw, models under different motion states are trained respectively. The trajectory of the training data set is divided into segments under different states at the command speed V=0. The tension and compression states of the feed shaft screw in each segment are judged. The residual error in each data segment is subtracted from the residual error of the first data group of the current data segment, and a residual error prediction model (trained machine learning model) relative to the tension state or compression state at the starting point of each segment is trained.
[0039] 2. The default tool setting point is the zero error point. Considering that when the machine tool feed axis worktable is at the tool setting point and the command speed V=0, the reverse clearance and deformation release are generated when the motion state of the machine tool feed axis is converted from the previous segment. When the motion state of the machine tool feed axis is not converted from the previous segment, the deformation continues to accumulate and no reverse clearance is generated. The training data sets with unchanged motion state and changed motion state at the command speed V=0 are divided into training sets of different models.
[0040] 3. Since the machine tool feed axis has high repeatability, when reaching the same position with the same motion state and speed, the residual error is equal. The motion state and conversion situation between each segment in the trajectory corresponding to the instruction position and the previous segment are judged, and an imaginary trajectory with the same motion state and conversion situation as the trajectory corresponding to the instruction position from the tool setting point to the starting point of each segment is planned. The residual error of the first instruction position of each segment is predicted through the trained machine learning model.
[0041] 4. The relative residual error at the starting point of each segment and the residual error of the first instruction position of each segment are predicted by the fusion parameterized model and machine learning. This improves the simulation accuracy of the machine tool feed axis and has good applicability. BRIEF DESCRIPTION OF THE DRAWINGS
[0042] Some embodiments of the present disclosure are described below with reference to the accompanying drawings, in which:
[0043] Figure 1 It is a flow chart of a method for predicting the position of a machine tool feed axis worktable provided in an embodiment of the present application.
[0044] Figure 2 It is a schematic diagram of the internal structure of a position prediction system for a machine tool feed axis worktable provided in an embodiment of the present application. DETAILED DESCRIPTION
[0045] It should be understood by those skilled in the art that the embodiments described below are only preferred embodiments of the present disclosure, and do not mean that the present disclosure can only be implemented through the preferred embodiments. The preferred embodiments are only used to explain the technical principles of the present disclosure, and are not used to limit the protection scope of the present disclosure. Based on the preferred embodiments provided by the present disclosure, all other embodiments obtained by ordinary technicians in this field without creative work should still fall within the protection scope of the present disclosure.
[0046] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.
[0047] The technical solution proposed in the embodiments of the present application is described in detail below with reference to the accompanying drawings.
[0048] The present application embodiment provides a method for predicting the position of a machine tool feed axis worktable, such as Figure 1 As shown, the method provided in the embodiment of the present application mainly includes the following steps:
[0049] Step 110, obtain the instruction information issued to the servo system corresponding to the machine tool feed axis worktable in the historical time period, input the instruction information into the preset parameterized simulation model to obtain the initial predicted position corresponding to each instruction information; collect the historical data corresponding to the machine tool feed axis worktable in the historical time period.
[0050] It should be noted that the instruction information includes the instruction position and instruction speed, and the historical data consists of several data groups, each of which includes at least: instruction information, initial predicted position and residual error, and the residual error is the difference between the actual workbench position collected and the initial predicted position.
[0051] Specifically, the command information issued to the servo system during the historical time period includes the command position, command speed, the simulated position of the feed axis worktable obtained by the parametric model simulation during the historical time period (the feature of machine learning); and the residual error (the label of machine learning) obtained by subtracting the actual position of the feed axis worktable collected during the historical time period from the simulated position of the feed axis worktable obtained by the parametric model, and is made into historical data for model training.
[0052] Step 120, taking the data group with instruction speed of 0 in the historical data as the breakpoint, divide the historical data of the historical time period into several data segments; subtract the residual error of the first data group of the current data segment from the residual error in each data group in each data segment, and determine the subtracted data segment as the training data segment.
[0053] It should be noted that a data segment is composed of data groups. The breakpoint data group can be placed at the head of the next data segment.
[0054] Step 130: determine the order of the training data segments based on the time sequence of the historical time periods; then determine the screw state conversion status corresponding to the current training data segment and the previous training data segment, and then divide the several training data segments into four training data sets; train the corresponding initial machine learning model through the training data set to obtain four trained machine learning models.
[0055] It should be noted that the training data segments with the earlier time sequence of the historical time segments are arranged in the order of the training data segments, and then arranged in order. The state of the lead screw is divided into a tension state and a compression state.
[0056] Among them, the screw state conversion conditions corresponding to the current training data segment and the previous training data segment are determined, and then the several training data segments are divided into four training data sets, which can be specifically:
[0057] Based on the information instructions corresponding to the current training data segment and the previous training data segment, the screw state change of the current training data segment is determined; wherein the screw state change is divided into a stretching-compression state, a stretching-stretching state, a compression-compression state, and a compression-stretching state; and then based on the four screw state changes, the training data segment is divided into corresponding training data sets.
[0058] Step 140, obtaining the trajectory to be predicted; inputting the instruction information to be predicted into a preset parameterized simulation model to obtain an initial predicted position of the trajectory to be predicted.
[0059] It should be noted that the specific training process involved in this step is an existing process, and this application does not limit the specific training process. The trajectory to be predicted is composed of the instruction information to be predicted, and the instruction information to be predicted is composed of the instruction position to be predicted and the instruction speed to be predicted.
[0060] Step 150: The predicted trajectory is divided into several predicted segments with the command speed of 0 as the breakpoint; the lead screw state conversion corresponding to each predicted segment of the predicted trajectory is obtained, and each predicted segment is input into the corresponding trained machine learning model to obtain the relative residual error of each predicted segment of the predicted trajectory; based on the tool setting point position, the first predicted command position of each predicted segment, and the maximum value of the predicted command speed in the current predicted segment, a polynomial interpolation method is used to plan the distance from the tool setting point position to the first predicted position of each predicted segment. A preset hypothetical trajectory of a predicted instruction position; wherein the preset hypothetical trajectory is composed of hypothetical instruction information, and the hypothetical instruction information is composed of a hypothetical instruction position and a hypothetical instruction speed; the screw state conversion situation of the first hypothetical instruction position of the preset hypothetical trajectory corresponding to the same to-be-predicted segment and the screw state conversion situation of the first to-be-predicted instruction position of the to-be-predicted trajectory are the same; the hypothetical instruction information of the preset hypothetical trajectory is input into a preset parameterized simulation model to obtain an initial predicted position of the preset hypothetical trajectory; the preset hypothetical trajectory is taken as a breakpoint at a command speed of 0 to obtain a plurality of hypothetical trajectory segments.
[0061] It should be noted that the preset parameterized simulation model is an existing model for predicting the initial predicted position based on the command information of the servo system, and this application uses it directly.
[0062] In addition, before obtaining the instruction information sent to the servo system corresponding to the feed axis worktable of the machine tool and inputting the instruction information into the preset parameterized simulation model, the method further includes:
[0063] Get the preset parameterized simulation model through the preset simulation software.
[0064] Step 160: obtain the state transition of the lead screw corresponding to each imaginary trajectory segment, and then input each imaginary trajectory segment into the corresponding trained machine learning model to obtain a preset error set corresponding to the preset imaginary trajectory; based on the specific number of imaginary trajectory segments, determine the residual error of the first instruction position of the current segment to be predicted.
[0065] It should be noted that the preset error set is composed of the preset errors corresponding to each hypothetical instruction information in the hypothetical trajectory segment.
[0066] Among them, based on the specific number of hypothetical trajectory segments, the residual error of the first command position of the current segment to be predicted is determined, which can be specifically:
[0067] When the specific number of hypothetical trajectory segments is 1, the preset error set consists of the preset errors of the unique hypothetical trajectory segment: {y1, y2, y3…yn}, and the preset error yn corresponding to the last hypothetical command position is determined as the residual error of the first command position of the current segment to be predicted;
[0068] When the specific number m of the hypothetical trajectory segments is greater than 1, the preset error set is composed of the preset errors of several hypothetical trajectory segments: {( , , … )、( , , … )、...( , , … )}, determine the sum of the preset errors corresponding to the last imaginary command position of each imaginary trajectory segment is the residual error of the first instruction position of the current segment to be predicted.
[0069] Step 170 , obtaining a final predicted position based on the initial predicted position of the trajectory to be predicted, the relative residual errors of each segment, and the residual error of the first command position of each segment to be predicted.
[0070] This step can be specifically as follows:
[0071] By formula:
[0072] ,
[0073] The final predicted position H is calculated; where h represents the initial predicted position, j represents the jth trajectory segment to be predicted, and Z represents the total number of trajectory segments to be predicted. represents the relative residual error of each segment to be predicted of the jth predicted trajectory, represents the residual error of the first command position of the jth trajectory segment to be predicted.
[0074] besides, Figure 2 A position prediction system for a machine tool feed axis table is provided in an embodiment of the present application. Figure 2 As shown, the system provided in the embodiment of the present application mainly includes:
[0075] The historical data acquisition module 210 is used to obtain the instruction information issued to the servo system corresponding to the machine tool feed axis worktable in the historical time period, and input the instruction information into the preset parameterized simulation model to obtain the initial predicted position corresponding to each instruction information; wherein the instruction information includes the instruction position and the instruction speed; collect the historical data corresponding to the machine tool feed axis worktable in the historical time period; wherein the historical data is composed of a plurality of data groups, each data group includes at least: instruction information, initial predicted position and residual error, and the residual error is the difference between the actual worktable position collected and the initial predicted position.
[0076] The training data acquisition module 220 is used to divide the historical data of the historical time period into several data segments with the data group with instruction speed of 0 in the historical data as the breakpoint; subtract the residual error of the first data group of the current data segment from the residual error in each data group in each data segment, and determine the subtracted data segment as the training data segment.
[0077] The data set division module 230 is used to determine the order between the training data segments based on the time sequence of the historical time periods; and then determine the screw state conversion corresponding to the current training data segment and the previous training data segment, and then divide the several training data segments into four training data sets; wherein the screw state is divided into a tension state and a compression state; train the corresponding initial machine learning model through the training data set to obtain four trained machine learning models.
[0078] The data set partitioning module 230 includes a partitioning unit,
[0079] Used to determine the state change of the lead screw of the current training data segment based on the information instructions corresponding to the current training data segment and the previous training data segment;
[0080] Among them, the state changes of the lead screw are divided into tension-compression state, tension-tension state, compression-compression state, and compression-tension state;
[0081] Then, based on the four screw state changes, the training data segments are divided into corresponding training data sets.
[0082] The predicted position acquisition module 240 is used to obtain the trajectory to be predicted; wherein the trajectory to be predicted is composed of the instruction information to be predicted, and the instruction information to be predicted is composed of the instruction position to be predicted and the instruction speed to be predicted; the instruction information to be predicted is input into the preset parameterized simulation model to obtain the initial predicted position of the trajectory to be predicted.
[0083] The hypothetical data acquisition module 250 is used to divide the issued to-be-predicted instruction information into a number of to-be-predicted segments with the instruction speed of 0 as the breakpoint of the to-be-predicted trajectory; obtain the state conversion of the lead screw corresponding to each to-be-predicted segment of the to-be-predicted trajectory, input each to-be-predicted segment into the corresponding trained machine learning model, and obtain the relative residual error of each to-be-predicted segment of the to-be-predicted trajectory; based on the tool setting point position, the first to-be-predicted instruction position of each to-be-predicted segment, and the maximum value of the to-be-predicted instruction speed in the current to-be-predicted segment, use the polynomial interpolation method to plan the distance from the tool setting point position to each to-be-predicted segment A preset hypothetical trajectory of the first to-be-predicted instruction position of a segment; wherein the preset hypothetical trajectory is composed of hypothetical instruction information, and the hypothetical instruction information is composed of hypothetical instruction position and hypothetical instruction speed; the screw state conversion situation of the first hypothetical instruction position of the preset hypothetical trajectory corresponding to the same to-be-predicted segment is the same as the screw state conversion situation of the first to-be-predicted instruction position of the to-be-predicted trajectory; the hypothetical instruction information of the preset hypothetical trajectory is input into a preset parameterized simulation model to obtain an initial predicted position of the preset hypothetical trajectory; the preset hypothetical trajectory is taken as a breakpoint at a command speed of 0 to obtain a plurality of hypothetical trajectory segments.
[0084] The residual error acquisition module 260 is used to obtain the screw state conversion corresponding to each imaginary trajectory segment, and then input each imaginary trajectory segment into the corresponding trained machine learning model to obtain a preset error set corresponding to the preset imaginary trajectory; wherein the preset error set is composed of the preset errors corresponding to each imaginary instruction information in the imaginary trajectory segment; based on the specific number of imaginary trajectory segments, determine the residual error of the first instruction position of the current segment to be predicted.
[0085] The residual error acquisition module 260 includes an error calculation unit,
[0086] When the specific number of hypothetical trajectory segments is 1, the preset error set is composed of the preset errors of the unique hypothetical trajectory segment: {y1, y2, y3…yn}, and the preset error yn corresponding to the last hypothetical command position is determined as the residual error of the first command position of the current segment to be predicted;
[0087] When the specific number m of the hypothetical trajectory segments is greater than 1, the preset error set is composed of the preset errors of several hypothetical trajectory segments: {( , , … )、( , , … )、...( , , … )}, determine the sum of the preset errors corresponding to the last imaginary command position of each imaginary trajectory segment is the residual error of the first instruction position of the current segment to be predicted.
[0088] The final position acquisition module 270 is used to obtain the final predicted position based on the initial predicted position of the trajectory to be predicted, the relative residual error of each segment and the residual error of the first command position of each segment to be predicted.
[0089] The final position acquisition module 270 includes a final position calculation unit, which is used to calculate the final position by the formula:
[0090] ,
[0091] The final predicted position H is calculated; where h represents the initial predicted position, j represents the jth trajectory segment to be predicted, and Z represents the total number of trajectory segments to be predicted. represents the relative residual error of each segment to be predicted of the jth predicted trajectory, represents the residual error of the first command position of the jth trajectory segment to be predicted.
[0092] In addition, an embodiment of the present application further provides a non-volatile computer storage medium on which executable instructions are stored. When the executable instructions are executed, a position prediction method for a machine tool feed axis worktable as described above is implemented.
[0093] So far, the technical solutions of the present disclosure have been described in combination with the above multiple embodiments, but it is easy for those skilled in the art to understand that the protection scope of the present disclosure is not limited to these specific embodiments. Without departing from the technical principles of the present disclosure, those skilled in the art can split and combine the technical solutions in the above-mentioned various embodiments, and can also make equivalent changes or replacements to the relevant technical features. Any changes, equivalent replacements, improvements, etc. made within the technical concept and / or technical principle of the present disclosure will fall within the protection scope of the present disclosure.
Claims
1. A method for predicting the position of a machine tool feed axis table, characterized in that: The method comprises: Obtaining command information issued to the servo system corresponding to the feed axis worktable of the machine tool in the historical time period, and inputting the command information into a preset parameterized simulation model to obtain an initial predicted position corresponding to each command information; wherein the command information includes a command position and a command speed; Collect historical data corresponding to the feed axis worktable of the machine tool within a historical time period; wherein the historical data is composed of a plurality of data groups, each of which includes at least: instruction information, an initial predicted position and a residual error, and the residual error is the difference between the actual worktable position collected and the initial predicted position; Taking the data group where the instruction speed is 0 in the historical data as the breakpoint, the historical data of the historical time period is divided into several data segments; subtracting the residual error of the first data group of the current data segment from the residual error in each data group in each data segment, and determining the subtracted data segment as the training data segment; Based on the time sequence of the historical time periods, the order between the training data segments is determined; then the state conversion of the lead screw corresponding to the current training data segment and the previous training data segment is determined, and then several training data segments are divided into four training data sets; wherein the lead screw state is divided into a tension state and a compression state; the corresponding initial machine learning model is trained through the training data set to obtain four trained machine learning models; Acquire a trajectory to be predicted; wherein the trajectory to be predicted is composed of instruction information to be predicted, and the instruction information to be predicted is composed of an instruction position to be predicted and an instruction speed to be predicted; Inputting the command information to be predicted into a preset parameterized simulation model to obtain an initial predicted position of the trajectory to be predicted; The predicted trajectory is divided into several predicted segments with the command speed of 0 as the breakpoint; the screw state conversion corresponding to each predicted segment of the predicted trajectory is obtained, and each predicted segment is input into the corresponding trained machine learning model to obtain the relative residual error of each predicted segment of the predicted trajectory; based on the tool setting point position, the first predicted command position of each predicted segment, and the maximum value of the predicted command speed in the current predicted segment, a preset hypothetical trajectory from the tool setting point position to the first predicted command position of each predicted segment is planned using the polynomial interpolation method; wherein the preset hypothetical trajectory is composed of hypothetical command information, and the hypothetical command information is composed of hypothetical command position and hypothetical command speed; the screw state conversion of the first hypothetical command position of the preset hypothetical trajectory corresponding to the same predicted segment and the screw state conversion of the first predicted command position of the predicted trajectory are the same; Inputting the hypothetical command information of the preset hypothetical trajectory into the preset parameterized simulation model to obtain the initial predicted position of the preset hypothetical trajectory; taking the preset hypothetical trajectory as a breakpoint at a command speed of 0 to obtain a plurality of hypothetical trajectory segments; Obtain the state conversion of the lead screw corresponding to each imaginary trajectory segment, and then input each imaginary trajectory segment into the corresponding trained machine learning model to obtain a preset error set corresponding to the preset imaginary trajectory; wherein the preset error set is composed of preset errors corresponding to each imaginary instruction information in the imaginary trajectory segment; Based on the specific number of hypothetical trajectory segments, determining a residual error of a first command position of a current segment to be predicted; The final predicted position is obtained based on the initial predicted position of the trajectory to be predicted, the relative residual errors of each segment and the residual errors of the first command position of each segment to be predicted.
2. The position prediction method of the machine tool feed axis table according to claim 1, characterized in that: Based on the specific number of hypothetical trajectory segments, the residual error of the first command position of the current segment to be predicted is determined, specifically including: When the specific number of hypothetical trajectory segments is 1, the preset error set consists of the preset errors of the unique hypothetical trajectory segment: {y1, y2, y3…yn}, and the preset error yn corresponding to the last hypothetical command position is determined as the residual error Q of the first command position of the current segment to be predicted; When the specific number m of the hypothetical trajectory segments is greater than 1, the preset error set is composed of the preset errors of several hypothetical trajectory segments: {( , , … )、( , , … )、...( , , … )}, determine the sum of the preset errors corresponding to the last imaginary command position of each imaginary trajectory segment is the residual error Q of the first command position of the current segment to be predicted.
3. The position prediction method of the machine tool feed axis table according to claim 1, characterized in that: Determine the screw state conversion status corresponding to the current training data segment and the previous training data segment, and then divide the several training data segments into four training data sets, specifically including: Determine the state change of the lead screw of the current training data segment based on the information instructions corresponding to the current training data segment and the previous training data segment; Among them, the state changes of the lead screw are divided into tension-compression state, tension-tension state, compression-compression state, and compression-tension state; Then, based on the four screw state changes, the training data segments are divided into corresponding training data sets.
4. The position prediction method of the machine tool feed axis table according to claim 1, characterized in that: Based on the initial predicted position of the trajectory to be predicted, the relative residual error of each segment and the residual error of the first command position of each segment to be predicted, the final predicted position is obtained, specifically including: By formula: ， The final predicted position H is calculated; where h represents the initial predicted position, j represents the jth trajectory segment to be predicted, and Z represents the total number of trajectory segments to be predicted. represents the relative residual error of each segment to be predicted of the jth predicted trajectory, represents the residual error of the first command position of the jth trajectory segment to be predicted.
5. A position prediction system for a machine tool feed axis table, characterized in that: The system comprises: A historical data acquisition module is used to obtain the instruction information issued to the servo system corresponding to the machine tool feed axis worktable in the historical time period, and input the instruction information into a preset parameterized simulation model to obtain the initial predicted position corresponding to each instruction information; wherein the instruction information includes the instruction position and the instruction speed; collect the historical data corresponding to the machine tool feed axis worktable in the historical time period; wherein the historical data is composed of a plurality of data groups, each of which at least includes: instruction information, initial predicted position and residual error, and the residual error is the difference between the actual worktable position collected and the initial predicted position; The training data acquisition module is used to divide the historical data of the historical time period into several data segments by taking the data group where the instruction speed is 0 in the historical data as the breakpoint; subtract the residual error of the first data group of the current data segment from the residual error in each data group in each data segment, and determine the subtracted data segment as the training data segment; A data set division module is used to determine the order between training data segments based on the time sequence of historical time segments; further determine the screw state conversion status corresponding to the current training data segment and the previous training data segment, and further divide several training data segments into four training data sets; wherein the screw state is divided into a tension state and a compression state; train the corresponding initial machine learning model through the training data set to obtain four trained machine learning models; A predicted position acquisition module is used to acquire a trajectory to be predicted; wherein the trajectory to be predicted is composed of instruction information to be predicted, and the instruction information to be predicted is composed of an instruction position to be predicted and an instruction speed to be predicted; the instruction information to be predicted is input into a preset parameterized simulation model to obtain an initial predicted position of the trajectory to be predicted; The hypothetical data acquisition module is used to divide the issued to-be-predicted instruction information into several to-be-predicted segments with the instruction speed of 0 as the breakpoint of the to-be-predicted trajectory; obtain the state conversion of the lead screw corresponding to each to-be-predicted segment of the to-be-predicted trajectory, input each to-be-predicted segment into the corresponding trained machine learning model, and obtain the relative residual error of each to-be-predicted segment of the to-be-predicted trajectory; based on the tool setting point position, the first to-be-predicted instruction position of each to-be-predicted segment, and the maximum value of the to-be-predicted instruction speed in the current to-be-predicted segment, the polynomial interpolation method is used to plan the distance from the tool setting point position to each to-be-predicted segment. A preset hypothetical trajectory of a first command position to be predicted; wherein the preset hypothetical trajectory is composed of hypothetical command information, and the hypothetical command information is composed of hypothetical command position and hypothetical command speed; the state transition of the lead screw at the first hypothetical command position of the preset hypothetical trajectory corresponding to the same segment to be predicted is the same as the state transition of the lead screw at the first command position to be predicted of the trajectory to be predicted; the hypothetical command information of the preset hypothetical trajectory is input into a preset parameterized simulation model to obtain an initial predicted position of the preset hypothetical trajectory; the preset hypothetical trajectory is taken as a breakpoint at a command speed of 0 to obtain a plurality of hypothetical trajectory segments; A residual error acquisition module is used to obtain the state conversion of the lead screw corresponding to each imaginary trajectory segment, and then input each imaginary trajectory segment into the corresponding trained machine learning model to obtain a preset error set corresponding to the preset imaginary trajectory; wherein the preset error set is composed of preset errors corresponding to each imaginary instruction information in the imaginary trajectory segment; based on the specific number of imaginary trajectory segments, determine the residual error of the first instruction position of the current segment to be predicted; The final position acquisition module is used to obtain the final predicted position based on the initial predicted position of the trajectory to be predicted, the relative residual error of each segment and the residual error of the first command position of each segment to be predicted.
6. The position prediction system of the machine tool feed axis table according to claim 5, characterized in that: The residual error acquisition module includes an error calculation unit, When the specific number of hypothetical trajectory segments is 1, the preset error set is composed of the preset errors of the unique hypothetical trajectory segment: {y1, y2, y3…yn}, and the preset error yn corresponding to the last hypothetical command position is determined as the residual error Q of the first command position of the current segment to be predicted; When the specific number m of the hypothetical trajectory segments is greater than 1, the preset error set is composed of the preset errors of several hypothetical trajectory segments: {( , , … )、( , , … )、...( , , … )}, determine the sum of the preset errors corresponding to the last imaginary command position of each imaginary trajectory segment is the residual error Q of the first command position of the current segment to be predicted.
7. The position prediction system of the machine tool feed axis table according to claim 5, characterized in that: The data set partitioning module includes a partitioning unit, Used to determine the state change of the lead screw of the current training data segment based on the information instructions corresponding to the current training data segment and the previous training data segment; Among them, the state changes of the lead screw are divided into tension-compression state, tension-tension state, compression-compression state, and compression-tension state; Then, based on the four screw state changes, the training data segments are divided into corresponding training data sets.
8. The position prediction system of the machine tool feed axis table according to claim 5, characterized in that: The final position acquisition module includes a final position calculation unit, Used by the formula: ， The final predicted position H is calculated; where h represents the initial predicted position, j represents the jth trajectory segment to be predicted, and Z represents the total number of trajectory segments to be predicted. represents the relative residual error of each segment to be predicted of the jth predicted trajectory, represents the residual error of the first command position of the jth trajectory segment to be predicted.
9. A non-volatile computer storage medium, characterized in that: Computer instructions are stored thereon, and when the computer instructions are executed, a position prediction method for a machine tool feed axis worktable as described in any one of claims 1 to 4 is implemented.