Profile cutting machine tool vibration suppression method, control device and cutting machine

Abstract

The embodiment of the invention provides a profile cutting machine tool vibration suppression method, a control device and a cutting machine. The method comprises the steps that S1, tool vibration historical data are collected, and a vibration data set is made; s2, constructing a mixed vibration prediction model, wherein the mixed vibration prediction model comprises an LSTM model and a tool-workpiece contact physical equation; s3, training the LSTM model by using the vibration data set to obtain a vibration prediction model, and connecting the vibration prediction model with the tool-workpiece contact physical equation in parallel; s4, collecting vibration data of the cutter in real time, and inputting the vibration data into the mixed vibration prediction model to obtain a prediction value; and S5, a vibration compensation signal is generated based on the predicted value, and the vibration compensation signal is used for carrying out vibration suppression on the tool. The method is based on the hybrid prediction model, vibration characteristics can be accurately captured, a vibration compensation signal is generated, and effective vibration suppression can be carried out on the cutter of the profile cutting machine.

Description

Technical Field

[0001] The present application relates to the field of vibration suppression, and particularly to a method for suppressing tool vibration of a profile cutting machine, a control device, and a cutting machine. Background Art

[0002] Vibration suppression of the tool of a profile cutting machine is crucial for improving machining accuracy and process stability. The vibration of the cutting tool not only increases the surface roughness of the workpiece and the number of burrs, but also accelerates tool wear, significantly increasing the machining cost. In fields such as aerospace and new energy vehicles, thin-walled profiles and composite materials have extremely high requirements for machining accuracy, and micron-level vibration errors can cause workpiece rejection. Traditional methods are difficult to meet the high-precision machining requirements due to a single prediction model and a lagging compensation strategy.

[0003] In the prior art, a pure data-driven model relies on a large amount of labeled data and lacks physical interpretability, while a pure physical model cannot adapt to dynamic working conditions. At the same time, traditional feedback control has problems of phase lag and amplitude mismatch, resulting in low efficiency in suppressing high-frequency vibration. There is an urgent need for an adaptive vibration suppression method that combines physical mechanisms and data characteristics. Summary of the Invention

[0004] The present application provides a method for suppressing tool vibration of a profile cutting machine, a control device, and a cutting machine. This method combines the time series learning ability of the LSTM model with the mechanism constraints of the tool-workpiece contact physical equation, which can not only capture the characteristics in vibration data but also constrain mechanism factors such as cutting force and material properties through physical equations, thereby improving the generalization ability of the prediction model under complex working conditions.

[0005] In a first aspect, a method for suppressing tool vibration of a profile cutting machine is provided. The method includes:

[0006] S1: Collect historical tool vibration data and make a vibration data set;

[0007] S2: Construct a hybrid vibration prediction model, where the hybrid vibration prediction model includes an LSTM model and a tool-workpiece contact physical equation;

[0008] S3: Use the vibration data set to train the LSTM model to obtain a vibration prediction model, and the vibration prediction model is connected in parallel with the tool-workpiece contact physical equation;

[0009] S4: Real-time collect the vibration data of the tool and input it into the hybrid vibration prediction model to obtain a predicted value;

[0010] S5: Generate a vibration compensation signal based on the predicted value, and the vibration compensation signal is generated for suppressing the vibration of the tool.

[0011] It should be understood that this method combines data-driven LSTM time series modeling with the physical mechanism constraints of the tool-workpiece contact process. On the one hand, by utilizing the adaptive learning ability of LSTM, it solves the problem that physical models are difficult to accurately describe dynamic nonlinear factors such as tool wear and chip morphology changes. On the other hand, prior knowledge such as tool stiffness and workpiece material properties is embedded into the model through physical equations, solving complex nonlinear effects that cannot be described by pure physical models, improving the generalization ability of the model, and effectively suppressing the vibration of the tool of the profile cutting machine.

[0012] Combined with the first aspect, in some implementation manners of the first aspect, the state equation of the hybrid vibration prediction model is:

[0013]

[0014] Where K c is the contact stiffness matrix, F t is the real-time cutting force, and α is the physical coupling coefficient.

[0015] Combined with the first aspect, in some implementation manners of the first aspect, the calculation formula of the contact stiffness matrix is:

[0016]

[0017] Where E is the average elastic modulus of the material, v is the Poisson's ratio of the material, T is the temperature of the contact area, w(y) is the equivalent contact width, and β is the softening coefficient.

[0018] It should be understood that the contact stiffness matrix proposed in this application dynamically calculates the tool-workpiece contact stiffness through the material elastic modulus E, Poisson's ratio v, equivalent contact width w(y), and temperature correction term tanh(β, T). This matrix can reflect the material softening effect and contact area change during the cutting process, combine with the LSTM model to predict the vibration trend, and form a hybrid prediction mechanism of physical and data collaboration. Compared with the traditional fixed stiffness model, it can improve the prediction accuracy.

[0019] It should be understood that α is the physical coupling coefficient, and its value range is [0, 1]. When α is 0, the vibration prediction is completely performed by the LSTM model; when α is 1, the vibration prediction is completely performed by the tool-workpiece contact physical equation.

[0020] It should also be understood that the equivalent contact width w(y) is related to the real-time cutting force F t and accurately depicts the actual contact area change between the tool and the workpiece by measuring the real-time cutting force F t This application of the present invention accurately depicts the actual contact area change between the tool and the workpiece, avoiding the error of the traditional uniform contact assumption.

[0021] Combined with the first aspect, in some implementation manners of the first aspect, the step S1 includes:

[0022] S101: Collect the vibration acceleration of the tool for capturing low-frequency vibration characteristics;

[0023] S102: Collect the high-frequency vibration signal of the tool for detecting the microscopic damage of the tool;

[0024] S103: Perform wavelet packet decomposition on the collected high-frequency vibration signal to filter high-frequency noise.

[0025] It should be understood that wavelet packet decomposition is an advanced signal processing technology. By recursively dividing the signal into finer time-frequency subbands through multi-level tree decomposition, it can capture both high-frequency transient characteristics and low-frequency steady-state information simultaneously. In this application, this method is applied to high-frequency signal processing, enabling the model to quickly identify high-frequency vibration sources and microscopic damage, thereby enhancing the real-time performance and accuracy of vibration prediction.

[0026] Combined with the first aspect, in some implementation manners of the first aspect, the step S5 includes:

[0027] S501: Obtain the predicted value of the hybrid vibration prediction model;

[0028] S502: Perform a 180° phase shift on the predicted value to generate an anti-phase signal;

[0029] S503: Calibrate and obtain the compensation displacement based on the anti-phase signal and the physical actuator characteristics;

[0030] S504: Convert the compensation displacement into a drive voltage by the physical actuator.

[0031] Combined with the first aspect, in some implementation manners of the first aspect, the method further includes online incremental training: after each cutting is completed, update the parameters in the LSTM model for the newly collected data based on the sliding window algorithm.

[0032] Combined with the first aspect, in some implementation manners of the first aspect, the sliding window algorithm includes:

[0033] S601: Window initialization, set the initial length of the window and the sliding step size;

[0034] S602: After each cutting cycle is completed, the window slides once for data update;

[0035] S603: Update the LSTM parameters based on the data within the window.

[0036] It should be understood that the sliding window mechanism maintains the temporal correlation by retaining the recent vibration data and dynamically updating the data, avoiding the model rigidity caused by the accumulation of historical data.

[0037] In combination with the first aspect, in some implementations of the first aspect, a PID controller is connected in series at the output end of the hybrid vibration prediction model, and the PID controller is used to adjust the predicted value in real time.

[0038] It should be understood that the main purpose of connecting the PID controller in series at the output end of the model is to adjust the prediction or output signal of the model in real time to optimize the dynamic response of the system and improve the control accuracy.

[0039] In a second aspect, a control device is provided, which is characterized by including a processor and a memory. The processor is coupled to the memory. The memory is used to store computer programs or instructions, and the processor is used to execute the computer programs or instructions in the memory, so that any implementation method described in the first aspect is executed.

[0040] In a third aspect, a cutting machine is provided, and the cutting machine includes the control device described in the second aspect. Description of the Drawings

[0041] Figure 1 It is a flowchart of an implementation method for suppressing the vibration of the tool of a profile cutting machine provided by an embodiment of the present application.

[0042] Figure 2 It is a flowchart of an implementation method for collecting the historical data of the tool vibration provided by an embodiment of the present application.

[0043] Figure 3 It is a flowchart of an implementation method for generating a vibration compensation signal provided by an embodiment of the present application.

[0044] Figure 4 It is a flowchart of an implementation method for parameter update based on a sliding window provided by an embodiment of the present application. Detailed Embodiments

[0045] The terms used in the following embodiments are only for the purpose of describing specific embodiments, and are not intended to limit the present application. As used in the specification and appended claims of the present application, the singular forms "a", "one kind", "", "the above", "the" and "this" are also intended to include the expression forms such as "one or more", unless there is a clear indication to the contrary in the context. It should also be understood that in the following embodiments of the present application, "at least one" and "one or more" mean one, two or more than two. The term "and / or" is used to describe the association relationship of associated objects, indicating that three relationships can exist; for example, A and / or B can represent: A exists alone, A and B exist simultaneously, and B exists alone, where A and B can be singular or plural. The character " / " generally represents an "or" relationship between the associated objects before and after.

[0046] References to "one embodiment" or "some embodiments" etc. described in this specification mean that a particular feature, structure, or characteristic described in connection with the embodiment is included in one or more embodiments of the present application. Thus, statements such as "in one embodiment", "in some embodiments", "in other some embodiments", "in still other embodiments", etc. that appear in different places in this specification do not necessarily all refer to the same embodiment, but mean "one or more but not all embodiments", unless otherwise specifically emphasized. The terms "comprising", "including", "having" and their variants all mean "including but not limited to", unless otherwise specifically emphasized.

[0047] Vibration suppression of the tool of a profile cutting machine plays a crucial role in ensuring machining quality and production efficiency. Tool vibration not only causes an increase in the surface roughness of the workpiece and the generation of burrs, but also accelerates the tool wear rate, thus significantly driving up the resource consumption cost in the production process.

[0048] The embodiments of the present application provide a method for suppressing tool vibration of a profile cutting machine, a control device, and a cutting machine. This method constructs a hybrid prediction model by integrating a data-driven LSTM model and a tool-workpiece contact physical equation, which can accurately capture vibration characteristics and generate vibration compensation signals, and can effectively suppress the vibration of the tool of the profile cutting machine.

[0049] Figure 1 FIG. is a flowchart for implementing a method for suppressing tool vibration of a profile cutting machine provided by the embodiments of the present application. In some examples, the method includes:

[0050] S1: Collect historical tool vibration data and make a vibration data set;

[0051] S2: Construct a hybrid vibration prediction model, where the hybrid vibration prediction model includes an LSTM model and a tool-workpiece contact physical equation;

[0052] S3: Train the LSTM model using the vibration data set to obtain a vibration prediction model, and the vibration prediction model is connected in parallel with the tool-workpiece contact physical equation;

[0053] S4: Real-time collect the vibration data of the tool and input it into the hybrid vibration prediction model to obtain a predicted value;

[0054] S5: Generate a vibration compensation signal based on the predicted value, and the vibration compensation signal is generated for suppressing the vibration of the tool.

[0055] In some examples, the state equation of the hybrid vibration prediction model is:

[0056]

[0057] Among them, K c is the contact stiffness matrix, F t is the real-time cutting force, and α is the physical coupling coefficient.

[0058] In some examples, the calculation formula of the contact stiffness matrix is:

[0059]

[0060] Among them, E is the average elastic modulus of the material, v is the Poisson's ratio of the material, T is the temperature of the contact area, w(y) is the equivalent contact width, and β is the softening coefficient.

[0061] In a possible implementation, a piezoresistive six-axis force sensor is used, installed at the connection flange between the tool holder and the spindle, to measure the three-axis cutting force (tangential, radial, axial) in real time. The sensor signal is converted to the tool coordinate system through coordinate transformation, and the centrifugal force interference is dynamically calibrated in combination with the spindle speed, and finally the accurate F t real-time cutting force data is output.

[0062] In some examples, step S1 includes:

[0063] S101: Collect the vibration acceleration of the tool to capture low-frequency vibration characteristics;

[0064] S102: Collect the high-frequency vibration signal of the tool to detect the microscopic damage of the tool;

[0065] S103: Perform wavelet packet decomposition on the collected high-frequency vibration signal to filter high-frequency noise.

[0066] In a possible implementation, a triaxial MEMS acceleration sensor is integrated at the tool holder to directly measure the vibration acceleration in the X / Y / Z axes at a specific sampling rate to obtain low-frequency vibration characteristics; a piezoelectric acoustic emission sensor is used, installed at the end face of the tool spindle, to collect high-frequency stress wave signals at a specific high sampling rate to extract effective high-frequency vibration signals.

[0067] In a possible implementation, perform 6-layer wavelet packet decomposition on the high-frequency vibration signal, divide the 1MHz signal into 64 sub-bands, manually select 16 effective sub-bands in the range of 200 - 800kHz, calculate the energy proportion of each sub-band, remove the high-frequency noise sub-bands with an energy proportion lower than 5%, retain the key feature frequency bands, and filter the noise.

[0068] In some examples, the said step S5 includes:

[0069] S501: Obtain the predicted value of the hybrid vibration prediction model;

[0070] S502: Perform a 180° phase shift on the predicted value to generate an anti-phase signal;

[0071] S503: Obtain a compensation displacement based on the anti-phase signal and physical actuator characteristic calibration;

[0072] S504: The physical actuator converts the compensation displacement into a drive voltage.

[0073] In a possible implementation, the physical actuator is a piezoelectric ceramic. The sensitivity of the piezoelectric ceramic is determined through offline experiments, and the drive voltage is determined based on the material hardness and the anti-phase signal.

[0074] In some examples, the method further includes online incremental training: after each cutting is completed, the parameters in the LSTM model are updated based on the newly acquired data using a sliding window algorithm.

[0075] In some examples, the sliding window algorithm includes:

[0076] S601: Window initialization, setting the initial length and sliding step of the window;

[0077] S602: After each cutting cycle is completed, the window slides once for data update;

[0078] S603: Update the LSTM parameters based on the data within the window.

[0079] In a possible implementation, the data within the window is used to update the LSTM parameters by the online gradient descent method, and the elastic weight consolidation algorithm is combined to retain important historical knowledge, enabling the model to quickly adapt to dynamic working conditions such as tool wear and material switching, and improving the machining qualification rate.

[0080] In some examples, a PID controller is connected in series at the output end of the hybrid vibration prediction model, and the PID controller is used to perform real-time adjustment on the predicted value.

[0081] In a possible implementation, the PID controller is connected in series at the output end of the hybrid model, and it receives in real time the vibration displacement signal predicted by the model and the actual displacement error feedback by the sensor. A correction signal is generated through proportional-integral-differential operations and superimposed on the piezoelectric actuator drive voltage to achieve vibration suppression.

[0082] An embodiment of the present application provides a control device, which includes a processor and a memory. The processor is coupled to the memory. The memory is used to store computer programs or instructions, and the processor is used to execute the computer programs or instructions in the memory, so that any of the methods in the foregoing embodiments is executed.

[0083] An embodiment of the present application provides a cutting machine, and the cutting machine includes the control device as described above.

[0084] The above is only the preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the above embodiments. Any equivalent modification or change made by those of ordinary skill in the art according to the disclosure of the present invention shall be included in the protection scope recorded in the claims.

Claims

1. A method for suppressing tool vibration of a profile cutting machine, characterized in that, The method includes: S1: Collect the historical data of the tool vibration and create a vibration data set; S2: Construct a hybrid vibration prediction model, where the hybrid vibration prediction model includes an LSTM model and a tool-workpiece contact physical equation; S3: Use the vibration data set to train the LSTM model to obtain a vibration prediction model, and the vibration prediction model is connected in parallel with the tool-workpiece contact physical equation; S4: Collect the vibration data of the tool in real time and input it into the hybrid vibration prediction model to obtain a predicted value; S5: Generate a vibration compensation signal based on the predicted value, and the vibration compensation signal is generated for vibration suppression of the tool.

2. The method according to claim 1, wherein The state equation of the hybrid vibration prediction model is: Among them, K c is the contact stiffness matrix, F t is the real-time cutting force, and α is the physical coupling coefficient.

3. The method according to claim 2, characterized in that The calculation formula of the contact stiffness matrix is: Where, E is the average elastic modulus of the material, v is the Poisson's ratio of the material, T is the temperature of the contact area, w(y) is the equivalent contact width, and β is the softening coefficient.

4. The method according to claim 1, characterized in that The step S1 includes: S101: Collect the vibration acceleration of the tool to capture low-frequency vibration characteristics; S102: Collect the high-frequency vibration signal of the tool to detect the microscopic damage of the tool; S103: Perform wavelet packet decomposition on the collected high-frequency vibration signal to filter out high-frequency noise.

5. The method according to claim 1, characterized in that, The step S5 includes: S501: Obtain the predicted value of the hybrid vibration prediction model; S502: Perform a 180° phase shift on the predicted value to generate an anti-phase signal; S503: Obtain a compensation displacement based on the anti-phase signal and the calibration of the physical actuator characteristics; S504: The physical actuator converts the compensation displacement into a driving voltage.

6. The method according to claim 1, wherein The method further includes online incremental training: after each cutting is completed, update the parameters in the LSTM model based on the newly collected data using a sliding window algorithm.

7. The method according to claim 6, wherein The sliding window algorithm includes: S601: Initialize the window, set the initial length of the window and the sliding step size; S602: After each cutting cycle is completed, the window slides once for data update; S603: Update the LSTM parameters using the online gradient descent method based on the data within the window.

8. The method according to claim 1, characterized in that, A PID controller is connected in series at the output end of the hybrid vibration prediction model, and the PID controller is used to adjust the predicted value in real time.

9. A control device, characterized in that, It includes a processor and a memory, the processor is coupled with the memory, the memory is used to store computer programs or instructions, and the processor is used to execute the computer programs or instructions in the memory, so that the method according to any one of claims 1 to 8 is executed.

10. A cutting machine, characterized in that, The cutting machine includes the control device according to claim 9.

Images