Mechanism-data fusion numerical control machine tool feeding system following error prediction method

Abstract

The invention discloses a mechanism-data fusion numerical control machine tool feeding system following error prediction method. The method comprises the following steps: constructing a dynamic simulation model of a numerical control machine tool feeding system, and applying an instruction position to obtain actual and simulation positions and intermediate simulation data; and establishing a residual error prediction network, training the network by taking the intermediate simulation data as input and taking the residual error between the actual position and the simulation position as output to obtain the residual error prediction network driven by the simulation data, and further correcting the output position of the simulation model so as to perform following error prediction. By combining the dynamic simulation model and the residual prediction network driven by the simulation data, accurate prediction of the following error of the feeding system of the numerical control machine tool is achieved, reliable support is provided for high-precision machining of the numerical control machine tool, the production efficiency and the product quality are expected to be remarkably improved, potential machining deviation can be recognized in advance, and the machining precision of the numerical control machine tool is improved. Precision control and error compensation are achieved, and therefore the machining quality and the system stability are improved.

Description

Technical Field

[0001] The present invention relates to a method for predicting system following error, and particularly to a method for predicting following error of a feed system of a numerically controlled machine tool based on mechanism-data fusion. Background Art

[0002] The feed system of a numerically controlled machine tool is mainly responsible for controlling the movement of each axis of the machine tool to ensure the accurate position between the tool and the workpiece, which directly affects the machining accuracy, machining efficiency and stability of the numerically controlled machine tool, and determines the overall performance and machining quality of the machine tool. Due to advantages such as high rigidity, high efficiency, long service life, and large stroke, ball screw drive is widely used in the feed system of numerically controlled machine tools. During the operation of the feed system of a numerically controlled machine tool, the dynamic difference between the actual position and the command position of a single axis is called following error, and the single-axis following error is comprehensively reflected on the surface of the machined part to form contour error, which affects the machining contour accuracy of the part.

[0003] Predicting the following error of the feed axis and then implementing corresponding compensation strategies is one of the effective methods to improve the accuracy of the feed system of numerically controlled machine tools. Currently, there are mainly two methods for predicting the following error of the feed system: modeling and simulation based on mechanism analysis and error prediction based on data driving.

[0004] The modeling and simulation method based on mechanism analysis usually refers to modeling parts such as the control system and power system of the feed system of a numerically controlled machine tool through various modeling tools, and carrying out co-simulation through the interfaces between different software. In recent years, with the development of multi-domain modeling languages, some researchers have also used modeling languages to model and simulate the whole system, solving the problem of incompatible interfaces of multiple modeling software. However, the simulation modeling method has deficiencies in expressing some non-linear elements (such as friction, clearance, lubrication, etc.) of the machine tool feed system, and it is difficult to describe its dynamic performance completely and accurately.

[0005] The error prediction method based on data driving is an end-to-end modeling method, which uses the input-output data of the system to directly establish a response prediction model of the system. New-generation artificial intelligence algorithms are also increasingly applied in this field. Data driving essentially establishes an equivalent black box model of the modeling object. The modeling is simple, and the fitting effect for non-linear relationships in the modeling object is relatively good, and the simulation accuracy is high. However, the model has no theoretical support, the physical meaning is not clear enough, and the robustness of the model is not good enough. Summary of the Invention

[0006] To solve the problems existing in the background technology, the present invention provides a method for predicting the following error of the feed system of a numerically controlled machine tool based on mechanism-data fusion. The method of the present invention makes full use of the respective advantages of the mechanism analysis and data-driven methods, thereby bridging the gap between the two. Through this fusion, the following error of the feed system can be accurately predicted, providing a solid foundation for subsequent error compensation and other work.

[0007] The technical solution adopted by the present invention is as follows:

[0008] The method for predicting the following error of the feed system of a numerically controlled machine tool based on mechanism-data fusion of the present invention includes:

[0009] S1: According to the operating mechanism of the feed system of the numerically controlled machine tool, construct a dynamic simulation model of the feed system of the numerically controlled machine tool to achieve high-precision dynamic simulation of the feed system of the numerically controlled machine tool; apply the same command position to the actual feed system of the numerically controlled machine tool and its dynamic simulation model respectively to obtain the actual position, simulation position and intermediate simulation data.

[0010] S2: Establish a residual prediction network, use the intermediate simulation data as the input, and the residual between the actual position and the simulation position as the output to train the residual prediction network to obtain a trained residual prediction network driven by simulation data.

[0011] S3: Connect the residual prediction network driven by simulation data in series after the dynamic simulation model of the feed system of the numerically controlled machine tool, apply the target command position to the dynamic simulation model to obtain the simulation position to be corrected and the intermediate simulation data, input the intermediate simulation data to be corrected into the residual prediction network driven by simulation data, output the prediction result after processing, and predict the following error of the feed system of the numerically controlled machine tool according to the prediction result and the simulation position to be corrected.

[0012] In the step S1, the dynamic simulation model of the feed system of the numerically controlled machine tool includes a servo motor module, a drive control module and a mechanical transmission module connected through a physical signal converter, and the command position is transmitted sequentially in the drive control module, the servo motor module and the mechanical transmission module.

[0013] After establishing the dynamic simulation model, identify the electrical parameters of the servo motor module, the friction parameters of the mechanical transmission module, and the backlash of the ball screw nut pair to improve the accuracy of the simulation model parameter settings. Complete the online identification of the electrical parameters (inductance, resistance and magnetic flux) of the servo motor of the ball screw feed system, identify the friction parameters of the mechanical transmission module, design an experimental scheme to measure the backlash value of the ball screw nut pair, and improve the accuracy of the simulation model.

[0014] In the step S1 described above, the intermediate simulation data includes the simulation speed, simulation acceleration, simulation electromagnetic torque, and simulation quadrature-axis current of the workbench of the numerical control machine tool feed system.

[0015] In the step S2 described above, the residuals between the actual position and the simulation position, and the intermediate simulation data are arranged and corresponding in chronological order.

[0016] In the step S2 described above, the residual prediction network includes an input layer, a convolutional neural network (CNN) layer, an autoregressive long short-term memory (LSTM) layer, an attention mechanism module, a fully connected layer, and an output layer connected in sequence.

[0017] In the step S3 described above, the predicted result and the simulation position to be corrected are fused to obtain the compensated simulation position, which is used as the predicted value of the actual position, so as to obtain the predicted value of the following error between the target command position and the predicted value of the actual position, and realize the following error prediction of the numerical control machine tool feed system.

[0018] The beneficial effects of the present invention are:

[0019] By analyzing the mechanism of the numerical control machine tool feed system, building a dynamic simulation model, and combining a residual prediction network driven by simulation data, the present invention realizes the accurate prediction of the following error of the numerical control machine tool feed system, provides reliable support for the high-precision machining of numerical control machine tools, is expected to significantly improve production efficiency and product quality, helps to identify potential machining deviations in advance, realizes precision control and error compensation, and thus improves machining quality and system stability. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 is a framework diagram of the method of the present invention;

[0021] Figure 2 is a schematic diagram of the component modules of the numerical control machine tool feed system;

[0022] Figure 3 is a framework diagram of the residual prediction network driven by simulation data. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0023] Here, the exemplary embodiments will be described in detail, and the examples are shown in the drawings. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. On the contrary, they are only examples of devices and methods consistent with some aspects of the present invention as detailed in the appended claims.

[0024] The terms used in the present invention are for the purpose of describing specific embodiments only and are not intended to limit the present invention. The singular forms "a", "said", and "the" used in the present invention and the appended claims are also intended to include the plural forms unless the context clearly dictates otherwise. It should also be understood that the term "and / or" used herein refers to and encompasses any and all possible combinations of one or more of the associated listed items.

[0025] The present invention will be further described in detail below in conjunction with specific embodiments. As Figure 1 shown, the method for predicting the following error of the mechanism-data fusion numerical control machine tool feed system of the present invention is specifically as follows:

[0026] As Figure 1 shown, the method for predicting the following error of the mechanism-data fusion numerical control machine tool feed system of the present invention is specifically as follows:

[0027] First, according to the operating mechanism of the numerical control machine tool feed system, a dynamic simulation model of the numerical control machine tool feed system is constructed to achieve high-precision dynamic simulation of the numerical control machine tool feed system. The dynamic simulation model of the numerical control machine tool feed system includes a servo motor module, a drive control module, and a mechanical transmission module connected through a physical signal converter, and the command position is sequentially transmitted in the drive control module, the servo motor module, and the mechanical transmission module. The servo motor module is modeled by selecting a permanent magnet synchronous motor commonly used in the numerical control machine tool feed system, and the stator voltage equation, electromagnetic torque equation, and mechanical equation of the surface-mounted permanent magnet synchronous motor are established in the synchronous rotating coordinate system. The drive control module is modeled using a vector control method to construct a semi-closed-loop three-loop control system including a current loop, a speed loop, and a position loop. The mechanical transmission module is modeled with the common ball screw feed system in the numerical control machine as the research object, and components such as couplings, ball screws, worktables, and guide rails are simplified into a stiffness-damping-inertia system. The mechanical transmission module considers non-linear error factors such as inertia, flexibility, and friction that can be modeled. The contact force relationship of the ball screw nut pair is deduced, the coupling and the screw are regarded as flexible bodies, the rigid-flexible coupling effect is fully considered, and a "Coulomb + viscous + Stribeck" friction model is introduced to describe the friction behavior of the ball screw.

[0028] As Figure 2As shown in the figure, it is a dynamic simulation model of the feed system of a numerically controlled machine tool, which consists of three modules: a servo motor, drive control, and mechanical transmission. The main task of the drive control module is to receive the command position sent by the numerical control system and convert it into a control signal that can be executed by the servo motor. The servo motor module, as an energy conversion device, provides the necessary rotational speed and torque to enable the motor rotor to reach the predetermined position according to the specified motion. The mechanical transmission module is responsible for accurately converting the rotational motion of the servo motor rotor into the linear motion of the machine tool table. The simulation models of the servo motor module and the drive control module use a semi-closed-loop three-ring control negative feedback system to control the servo motor. From the inside out, they are the current loop, speed loop, and position loop. In the actual machine tool feed system, the relevant feedback signals are obtained through the motor encoder. A permanent magnet synchronous motor commonly used in the feed system of a numerically controlled machine tool is selected. On the premise of not affecting the control performance, a realistic and feasible mathematical model of the permanent magnet synchronous motor is established, and the mathematical model of the motor is decoupled by means of coordinate transformation. Each component of the mechanical transmission module (coupling, lead screw, nut, table) is simplified into a stiffness-damping-inertia system and connected through kinematic pairs such as sliding pairs, rotating pairs, and nut pairs. The different modules of electricity, control, and machinery are connected through a physical signal converter to complete the construction of the dynamic simulation model of the feed system of the numerically controlled machine tool.

[0029] After establishing the dynamic simulation model, the electrical parameters of the servo motor module, the friction parameters of the mechanical transmission module, and the backlash of the lead screw nut pair are identified to improve the accuracy of the simulation model parameter settings. The online identification of the electrical parameters (inductance, resistance, and magnetic flux) of the servo motor of the ball screw feed system is completed, the friction parameters of the mechanical transmission module are identified, and an experimental scheme is designed to measure the backlash value of the lead screw nut pair, improving the accuracy of the simulation model. The present invention proposes an online identification method for the electrical parameters of the feed system of a numerically controlled machine tool based on the recursive least squares method. Through experiments, it is verified that the friction characteristics of the feed system of the numerically controlled machine tool conform to the "Coulomb + viscous + Stribeck effect" friction model. The friction parameters are identified based on the genetic optimization algorithm, and the backlash value of the lead screw nut pair is measured through the reciprocating experiment of the table, realizing the accurate identification of the physical parameters of the simulation model of the feed system of the numerically controlled machine tool.

[0030] Then, the same command position is applied to the actual CNC machine tool feed system and its dynamic simulation model respectively to obtain the actual position, simulation position and intermediate simulation data. The intermediate simulation data includes the simulation speed, simulation acceleration, simulation electromagnetic torque and simulation quadrature axis current of the workbench of the CNC machine tool feed system; the residuals between the actual position and the simulation position, and the intermediate simulation data are arranged and corresponding in chronological order. During specific implementation, the same command position signal is applied to the actual CNC machine tool feed system and its corresponding simulation model, and the output signal data of actual observation and simulation are collected respectively; specifically: an ideal feed axis motion position signal is applied to the actual machine tool feed system and the simulation model, and relevant displacement output, speed output, simulation torque, simulation current and other data are collected through the grating scale installed on the actual machine tool and the virtual sensor in the simulation model. The collected actual response signal and simulation prediction signal data sequences are arranged in chronological order.

[0031] Then, a residual prediction network is established. The residual prediction network includes an input layer, a convolutional neural network (CNN) layer, an autoregressive long short-term memory (LSTM) layer, an attention mechanism module, a fully connected layer and an output layer, which are connected in sequence. The CNN layer performs one-dimensional convolution operations on the input multi-dimensional time step data to extract local features from the time series data of multiple input variables; the autoregressive LSTM layer can effectively capture and remember the long-term dependencies in the time series and comprehensively consider the influence of historical values; the attention mechanism module can make the model focus on more important feature inputs by assigning different weights and reduce the dependence on the predicted value of the previous time step. The CNN layer, the autoregressive LSTM layer, the attention mechanism module and the fully connected layer serve as the hidden layers of the residual prediction network. Using the intermediate simulation data as the input and the residual between the actual position and the simulation position as the output, the residual prediction network is trained to obtain a trained simulation data-driven residual prediction network. As Figure 3 shown, the multi-dimensional time step data {x1, x2, x3, …, x n} input to the input layer is input into the CNN layer, and convolution and pooling operations are performed in sequence, and then input into the autoregressive LSTM layer for processing; then input into the QKV attention mechanism module to assign different weights, and finally the prediction result y is output at the output layer after being processed by the fully connected layer.

[0032] Finally, a simulation data-driven residual prediction network is connected in series behind the dynamic simulation model of the CNC machine tool feeding system. Apply the target command position to the dynamic simulation model to obtain the simulation position to be corrected and intermediate simulation data. Input the intermediate simulation data to be corrected into the simulation data-driven residual prediction network. After processing, the prediction result is output. The prediction result and the simulation position to be corrected are fused to obtain the compensated simulation position, which is used as the actual position prediction value. Thus, the following error prediction value between the target command position and the actual position prediction value is obtained, realizing the following error prediction of the CNC machine tool feeding system.

[0033] During specific implementation, as Figure 1 shown, the same command position X0 is respectively input into the real CNC machine tool feeding system and its dynamic simulation model. After processing, the actual position X and the simulation position are obtained. Furthermore, the residual r between the actual position X and the simulation position is obtained. At the same time, the intermediate simulation data output by the simulation model is obtained, including the kinematic simulation quantity of the workbench, the simulated electromagnetic torque, and the simulated current, specifically the simulated speed, simulated acceleration, simulated electromagnetic torque, and simulated quadrature-axis current of the workbench. Then, the established residual prediction network is trained through the residual r and the intermediate simulation data to obtain the trained simulation data-driven residual prediction network. During actual prediction, the simulation data-driven residual prediction network is connected in series behind the dynamic simulation model. The target command position is input into the dynamic simulation model to obtain the intermediate simulation data and simulation position to be corrected, and then input into the simulation data-driven residual prediction network. After processing, the predicted residual r * is obtained. The predicted residual r * and the simulation position to be corrected are fused to obtain the actual position prediction value X * . Furthermore, the actual position prediction value X * and the following error prediction value e between the target command position * are obtained, realizing the following error prediction.

[0034] Those skilled in the art should understand that the embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the present application can adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, optical storage, etc.) containing computer-usable program code. The solutions in the embodiments of the present application can be implemented in various computer languages. The present application is described according to the flowcharts of the methods, systems, and computer program products of the embodiments of the present application.

[0035] Although the preferred embodiments of the present application have been described, those skilled in the art can make additional changes and modifications once they learn the basic creative concepts. Therefore, the present invention is intended to be construed as including the preferred embodiments and all changes and modifications that fall within the scope of the present application.

[0036] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the equivalent technology of the present invention, the present application is also intended to include these modifications and variations.

Claims

1. A mechanism-data fusion-based following error prediction method for the feed system of a numerically controlled machine tool, characterized in that Including: S1: Construct a dynamic simulation model of the CNC machine tool feed system, apply the same command position to the actual CNC machine tool feed system and its dynamic simulation model respectively, and obtain the actual position, simulation position and intermediate simulation data; S2: Establish a residual prediction network, use the intermediate simulation data as the input, and the residual between the actual position and the simulation position as the output to train the residual prediction network, and obtain a trained residual prediction network driven by simulation data; S3: Connect the residual prediction network driven by simulation data in series behind the dynamic simulation model of the CNC machine tool feed system, apply the target command position to the dynamic simulation model, obtain the simulation position to be corrected and the intermediate simulation data, input the intermediate simulation data to be corrected into the residual prediction network driven by simulation data, output the prediction result after processing, and predict the following error of the CNC machine tool feed system according to the prediction result and the simulation position to be corrected.

2. The method for predicting the following error of the feed system of a numerically controlled machine tool based on mechanism-data fusion according to claim 1, characterized in that: In the step S1, the dynamic simulation model of the CNC machine tool feed system includes a servo motor module, a drive control module and a mechanical transmission module connected through a physical signal converter, and the command position is transmitted sequentially in the drive control module, the servo motor module and the mechanical transmission module.

3. The method for predicting the following error of the feed system of a numerically controlled machine tool based on mechanism-data fusion according to claim 1, characterized in that: In the step S1, the intermediate simulation data includes the simulation speed, simulation acceleration, simulation electromagnetic torque and simulation quadrature axis current of the workbench of the CNC machine tool feed system.

4. The method for predicting the following error of the feed system of the numerically controlled machine tool with mechanism-data fusion according to claim 1, characterized in that: In the step S2, the residuals between the actual position and the simulation position and the intermediate simulation data are arranged and corresponding in chronological order.

5. The method for predicting the following error of the feed system of a numerically controlled machine tool based on mechanism-data fusion according to claim 1, characterized in that: In the step S2, the residual prediction network includes an input layer, a convolutional neural network CNN layer, an autoregressive long short-term memory LSTM layer, an attention mechanism module, a fully connected layer and an output layer connected in sequence.

6. The mechanism-data fusion-based following error prediction method for the feed system of a numerically controlled machine tool according to claim 1, wherein: In the step S3, the prediction result and the simulation position to be corrected are fused to obtain a compensated simulation position, which is used as the actual position prediction value, so as to obtain the following error prediction value between the target command position and the actual position prediction value, and realize the following error prediction of the CNC machine tool feed system.

7. An electronic device, characterized in that, Including: A memory and a processor coupled to each other, wherein the memory stores program data, and the processor calls the program data to execute the method according to any one of claims 1-6.

8. A computer-readable storage medium having program data stored thereon, characterized in that, When the program data is executed by the processor, the method according to any one of claims 1-6 is implemented.

Images