Machine tool vibration suppression optimization method and system

Abstract

This invention discloses a machine tool vibration suppression optimization method and system. The optimization method includes: generating a control strategy based on machining parameters; collecting vibration data from key parts of the CNC machine tool; inputting the vibration data and current working parameters into a vibration prediction model; determining whether the vibration prediction result of the vibration prediction model meets the machining requirements; if not, adjusting the machine tool working parameters in the control strategy using a reinforcement learning algorithm. This invention uses a combination of a vibration prediction model and a reinforcement learning algorithm to suppress vibration, achieving real-time vibration prediction and adaptive control. This effectively suppresses vibration, significantly improves machining accuracy, reduces equipment wear and failures, and extends the machine tool's service life.

Description

Technical Field

[0001] This invention belongs to the field of CNC machine tools, and in particular relates to a machine tool vibration suppression optimization method and system. Background Technology

[0002] Vibration is unavoidable during the machining process of CNC machine tools. The generation of machine tool vibration mainly stems from factors such as cutting force, the structural rigidity of machine tool components, and the external environment. Machine tool vibration not only affects the surface quality of the workpiece but can also lead to tool damage and increased machining errors. Machine tool vibration suppression is an important means to improve machining accuracy, extend machine tool life, and ensure machining quality, and it is of great significance in modern CNC machine tools and high-precision machining. However, traditional vibration suppression methods mainly rely on mechanical design and human experience, which are insufficient to cope with complex and variable vibration environments. Therefore, it is necessary to propose a new machine tool vibration suppression method. Summary of the Invention

[0003] In view of the shortcomings of the prior art, the technical problem to be solved by the present invention is to provide a machine tool vibration suppression optimization method and system.

[0004] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0005] A method for optimizing machine tool vibration suppression includes the following steps:

[0006] S100: Generate a control strategy based on the machining parameters, and generate real-time control signals based on the control strategy to control the working parameters of the machine tool in each operating cycle;

[0007] S200: Collect vibration data of key parts of the CNC machine tool during operation;

[0008] S300: Input the vibration data and the machine tool's current operating parameters into the vibration prediction model to predict subsequent vibration results;

[0009] S400: Determine whether the vibration prediction results predicted by the vibration prediction model meet the processing requirements. If not, proceed to step S500; otherwise, return to step S200.

[0010] S500: Based on the vibration prediction results from the vibration prediction model, a reinforcement learning algorithm is used to adjust the machine tool operating parameters in the control strategy.

[0011] Furthermore, the key components of the machine tool include the spindle, cutting tool, and worktable of the CNC machine tool; the working parameters include cutting speed, cutting acceleration, feed rate, depth of cut, spindle speed, tool type, and cutting path.

[0012] Furthermore, the method for collecting vibration data from key parts of a machine tool includes the following steps:

[0013] S210. Vibration acquisition devices are installed at key parts of the machine tool to collect vibration signals from the key parts of the machine tool;

[0014] S220. Denoise the vibration signal to eliminate high-frequency and low-frequency noise;

[0015] S230. Extract the frequency domain and time domain features of the vibration signal to obtain vibration data.

[0016] Furthermore, the method for obtaining the vibration prediction model includes the following steps:

[0017] S310. Determine a machine learning model or deep learning model as the base model for the vibration prediction model;

[0018] S320. Acquire historical vibration data of the machine tool, determine the normal state data and abnormal state data in the historical vibration data according to the vibration range when the machine tool is working normally, and mark the vibration data of the reverse point, start point and stop point as well as the abnormal state data.

[0019] S330. The basic model is trained using the labeled vibration data to obtain the vibration prediction model.

[0020] Furthermore, a Long Short-Term Memory (LSTM) network is used as the base model.

[0021] Furthermore, the reinforcement learning algorithm adopts the Q-learning algorithm, with the current machine tool operating parameter values ​​and vibration data as the state S of the Q-learning algorithm, the adjustment of the machine tool operating parameters as the action a of the Q-learning algorithm, and the reduction of machine tool vibration and / or the optimization of machine tool processing performance as the reward R of the Q-learning algorithm.

[0022] Furthermore, the method for adjusting the machine tool operating parameters in the control strategy using reinforcement learning algorithms includes the following steps:

[0023] S510. Obtain the current processing parameters and the vibration prediction results from the vibration prediction model, and represent them as the current state s. t ;

[0024] S520. Based on the current state, select an action a as the current action a using the ∈-greedy strategy. t ;

[0025] S530. Adjust the machine tool's operating parameters in the control strategy to execute the current action. t And while performing the current action a tAfter the predetermined time, real-time vibration data and processing results are obtained as feedback.

[0026] S540. Calculate the reward value r based on the feedback results. t+1 ;

[0027] S550, Update the Q value of the current state and action.

[0028] S560. Determine if the preset termination condition is met. If the termination condition is met, return to step S200; otherwise, return to step S510. For the current state s... t After updating, the control strategy is adjusted again until the requirements are met.

[0029] Furthermore, in step S540, action a is performed. t The subsequent reward value r t+1 The calculation formula is as follows:

[0030]

[0031] Among them, a v The weighting coefficients for the vibration data values; V(t) represents the machine tool's value during the execution of action a. t The values ​​of the feedback vibration data; i is the index of the number of parameters involved in calculating the reward value, N is the total number of parameters involved in calculating the reward value, b i f is the weighting coefficient for the i-th parameter. i (t) represents the i-th parameter during the execution of action a. t The feedback value is calculated afterward. When the parameter is set to be as large as possible, its weighting coefficient is positive; when the parameter is set to be as small as possible, its weighting coefficient is negative.

[0032] Furthermore, in step S550, the formula for updating the Q value is as follows:

[0033]

[0034] Where, Q′(s t ,a t ) is in state s t Next, execute action a t The Q value, i.e., the current action a t The expected cumulative reward value of this strategy; Q(s) t ,a t ) represents the Q-value before the update; α is the learning rate, representing the weight of the influence of new information on the existing Q-value, with a range of 0 < α ≤ 1; r t+1 To perform action a at the current time step t tThe immediate reward obtained afterward; γ is a discount factor, representing the degree of influence of future rewards on the current decision, with a value range of 0≤γ≤1; For the next state S t+1 In the middle, all selectable actions a′ t+1 The action with the highest Q value.

[0035] A machine tool vibration suppression optimization system, including

[0036] The control strategy module is used to generate control strategies based on machining parameters and generate real-time control signals to the machine tool actuators based on the control strategies, thereby controlling the working parameters of the machine tool actuators in each operating cycle.

[0037] The data acquisition module is used to collect vibration data of key parts of the machine tool during operation.

[0038] A vibration prediction module, trained using a machine learning or deep learning model, is used to predict vibration results for subsequent operating cycles based on vibration data and the machine tool's operating parameters in the current operating cycle; and

[0039] The parameter optimization module is used to adjust the machine tool operating parameters in the control strategy using a reinforcement learning algorithm when the vibration results predicted by the vibration prediction module do not meet the processing requirements.

[0040] This invention employs a method combining vibration prediction model for early prediction and reinforcement learning algorithm for optimized control strategy to suppress vibration. This achieves real-time vibration prediction and adaptive control, effectively suppressing vibration, significantly improving machining accuracy, reducing equipment wear and malfunctions, and extending machine tool lifespan. Furthermore, by continuously executing reinforcement learning algorithms to optimize the control strategy during machining, the control strategy can be continuously refined, ultimately enabling the machine tool to achieve optimal machining performance while minimizing vibration. Attached Figure Description

[0041] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0042] Figure 1 This is a flowchart of an embodiment of the machine tool vibration suppression optimization method of the present invention.

[0043] Figure 2 This is a flowchart for collecting vibration data from key parts of a machine tool.

[0044] Figure 3 The flowchart for obtaining the vibration prediction model.

[0045] Figure 4 This is a flowchart illustrating how a reinforcement learning algorithm is used to adjust the control strategy.

[0046] Figure 5 This is a structural block diagram of an embodiment of the machine tool vibration suppression and optimization system of the present invention.

[0047] The diagrams in the instruction manual are labeled as follows:

[0048] Control strategy module - 100; Machine tool actuator - 200; Data acquisition module - 300; Vibration prediction module - 400; Parameter optimization module - 500. Detailed Implementation

[0049] The following specific examples illustrate the implementation of the present invention. The illustrations provided in the following embodiments are only schematic representations of the basic concept of the present invention. Unless otherwise specified, the following embodiments and features can be combined with each other.

[0050] Currently, vibration suppression mainly employs spindle vibration suppression technology, feed axis vibration suppression technology, and tool vibration suppression technology.

[0051] There are three main methods for spindle vibration suppression: variable speed cutting, spindle bearing preload control, and spindle system self-balancing control. However, all have significant drawbacks. Variable speed cutting uses periodic, continuous changes in cutting speed to avoid unstable cutting zones, thus suppressing vibration. Its disadvantages include the need for extensive machining tests to establish a system stability limit diagram. Changes to any part of the machining system—spindle, tool, fixture, or workpiece—will alter the stability limit diagram, requiring re-planning of chatter prediction data. Spindle bearing preload control increases preload at low speeds and high torques, and decreases preload at high speeds and low torques. Its disadvantage is that it can only fix the bearing preload for a single application scenario. Spindle system self-balancing control uses a balancing mechanism to self-balance the spindle during high-speed rotation. Its disadvantages include the complexity of the balancing mechanism, making widespread application difficult; generally, spindle balance is only calibrated before machining.

[0052] Vibration suppression of the feed axis employs methods that improve the structure of mechanical transmission components, primarily through the use of new linear motors, electromagnetic shielding of electrical cabinets, and servo tuning. Its disadvantages include the high time and effort required, and limited economic benefits.

[0053] Tool vibration suppression mainly employs methods such as improving tool holder materials, optimizing fixtures, and optimizing process parameters. However, these methods have the disadvantages of requiring significant time and effort and yielding limited economic benefits.

[0054] Please see Figure 1 , Figure 1This is a flowchart of an embodiment of the machine tool vibration suppression optimization method of the present invention. The machine tool vibration suppression optimization method of this embodiment includes the following steps:

[0055] The S100 CNC machine tool generates a control strategy based on the machining parameters, and then generates real-time control signals based on the control strategy to control the machine tool's working parameters in each operating cycle. The machining parameters are generally parameters such as cutting speed, feed rate, depth of cut, and machining path that are pre-input during machining.

[0056] S200: Collect vibration data of key parts of the machine tool during machine tool operation. Please refer to [link / reference]. Figure 2 Methods for collecting vibration data from key parts of a machine tool may include the following steps:

[0057] S210. Vibration acquisition devices such as accelerometers, displacement sensors, and strain gauges are installed at key parts of the machine tool to collect vibration signals from these key parts. These key parts generally include the machine tool spindle, cutting tool, and worktable (including the feed axis).

[0058] S220. Use filters (such as Kalman filters, bandpass filters, etc.) to denoise the collected vibration signals and eliminate high-frequency and low-frequency noise.

[0059] S230. The frequency domain and time domain features of the vibration signal are extracted using methods such as Fourier transform (FFT) and wavelet transform to obtain vibration data.

[0060] S300: Input the vibration data and the machine tool's current operating parameters into the vibration prediction model to predict subsequent vibration results. The machine tool's operating parameters generally include cutting speed, cutting acceleration, feed rate, depth of cut, spindle speed, tool type, and cutting path. The operating cycle is an artificially defined period for ease of control; its duration can be selected according to actual needs. Within one operating cycle, the control strategy can issue multiple real-time control signals. For example, 1ms can be used as an operating cycle; alternatively, 0.1ms or other durations can be chosen, as long as the final control effect is achieved.

[0061] Please see Figure 3 The method for obtaining a vibration prediction model through pre-training can be replaced by the following steps:

[0062] S310. Determine a machine learning model (such as Support Vector Machine (SVM) or Random Forest) or deep learning model (such as Convolutional Neural Network (CNN) or Long Short-Term Memory Network (LSTM)) as the base model for the vibration prediction model. In this embodiment, Long Short-Term Memory Network (LSTM) is used as the base model.

[0063] S320. Acquire historical vibration data of the machine tool. Based on the vibration range during normal machine operation, determine the normal and abnormal state data in the historical vibration data, and mark the vibration data at reversal points, start points, and stop points, as well as the abnormal state data. When machining a new workpiece, vibration data can be obtained through trial cutting of the workpiece as historical vibration data for the machine tool. During machining, significant vibrations generally occur at reversal points, start points, and stop points; therefore, these points are marked together with the abnormal state data during training. The remaining data is considered normal state data. Generally, normal and abnormal state data are distinguished based on the historical vibration range. For example, if historical data shows vibration values ​​between 0 and 1, then data within this range is considered normal state data, and data outside this range is considered abnormal state data.

[0064] S330. The labeled vibration data is used to train the basic model to obtain the vibration prediction model. Of course, during the actual processing of the workpiece, the vibration data and related working parameters during the processing can be used as historical vibration data to continue training the model after each cycle is completed, thereby continuously improving the prediction accuracy of the vibration prediction model.

[0065] S400: Determine whether the vibration prediction results predicted by the vibration prediction model meet the processing requirements. If not, proceed to step S500; otherwise, return to step S200.

[0066] Due to the need to consider processing costs, the number of trial cuts on the workpiece is relatively limited. Training the model solely based on the data from these trial cuts can only initially meet processing requirements and cannot achieve the optimal control strategy. Therefore, in the initial processing stage, step S500 is typically executed to continuously optimize the control strategy. As the number of workpiece processing cycles increases, the model is continuously trained using vibration data from each cycle, allowing the workpiece processing control strategy to gradually stabilize through continuous optimization. At this point, under normal conditions, further optimization is generally unnecessary, and the process directly returns to step S200 for monitoring only, allowing for timely intervention in case of unexpected processing issues. Of course, if the machine tool experiences a collision or other events that alter its performance, or if a new workpiece is processed, then optimization must be performed again.

[0067] S500: Based on the vibration prediction results from the vibration prediction model, reinforcement learning algorithms (such as Q-learning, deep Q-network DQN, etc.) are used to adjust the machine tool working parameters in the control strategy.

[0068] In this embodiment, the reinforcement learning algorithm employs the Q-learning algorithm. The current machine tool operating parameter values ​​and vibration data (including vibration data predicted by the vibration prediction model and vibration data fed back after the action is executed) are used as the state S of the Q-learning algorithm, and the adjustment of the machine tool operating parameters is used as the action a. For example, stepping up or down the value of a certain operating parameter can be considered as action a, or adjusting the cutting path can be considered as action a. The reduction of machine tool vibration and the optimization of machine tool machining performance are used as the reward R of the Q-learning algorithm. Alternatively, the reduction of machine tool vibration can be used as the reward R alone.

[0069] Please see Figure 4 When the reinforcement learning algorithm uses the Q-learning algorithm, this step may include the following sub-steps:

[0070] S510. Obtain the current processing parameters and the vibration prediction results from the vibration prediction model, and represent them as the current state s. t Therefore, based on the vibration prediction results, possible vibration situations can be addressed in advance, and subsequent vibrations can be reduced by adjusting the control strategy, thereby achieving the effect of vibration suppression.

[0071] S520. Based on the current state, select an action a as the current action a using the ∈-greedy strategy (i.e., the decaying greedy strategy). t The formula is as follows:

[0072]

[0073] in, Let ∈ be the action that maximizes the Q-value in the current state. ∈ is the probability of exploration, meaning a random action is selected with probability ∈ to explore new possibilities and prevent getting trapped in local optima; 0 < ∈ < 1. 1-∈ is the probability of exploitation, meaning a action with probability 1-∈ is selected to further optimize the control policy ∈-greedy policy. The value of ∈ can gradually decrease as the number of iterations increases.

[0074] S530. Adjust the machine tool's operating parameters in the control strategy to execute the current action. t And while performing the current action a t Real-time vibration data and processing results are obtained after a predetermined time as feedback. For example, for locations with significant vibration due to the machine tool's assembly or structure, vibration can be reduced by decreasing the cutting speed. If adjusting parameters such as the cutting speed still fails to resolve the issue, the cutting path can be adjusted to bypass the location and avoid excessive vibration.

[0075] S540. Calculate the reward value r based on the feedback results. t+1 Since the purpose of this embodiment is to suppress vibration, the feedback results will include vibration data. For example, action a can be set to be performed. t The subsequent reward value r t+1 The calculation formula is as follows:

[0076] r t+1 = -V(t)

[0077] Where V(t) represents the machine tool performing action a t The vibration data is then fed back. The vibration data can be obtained using vibration amplitude or a combination of vibration performance parameters such as vibration acceleration and displacement. Of course, in actual processing, there may be requirements not only for vibration but also for processing efficiency and other processing effects. In this case, processing efficiency and other processing effect performance can also be included in the bonus value r. t+1 The calculation. At this time, action a is executed. t The subsequent reward value r t+1 The calculation formula is as follows:

[0078]

[0079] Among them, a v The weighting coefficients for vibration data values; i is the index of the number of parameters involved in calculating the reward value, i = 0 indicates that performance aspects such as processing efficiency are not considered. N is the total number of parameters involved in calculating the reward value, b i f is the weighting coefficient for the i-th parameter. i (t) represents the i-th parameter during the execution of action a. t The feedback value afterwards.

[0080] In this embodiment, the parameters involved in calculating the reward value include machine tool vibration performance parameters, processing efficiency, and workpiece surface roughness; therefore, action a is executed. t The subsequent reward value r t+1 The calculation formula is as follows:

[0081] r t+1 =-a v V(t) + b1f1(t) + b2f2(t)

[0082] Where b1 is the weighting coefficient of the machine tool's processing efficiency, and f1(t) is the machine tool's efficiency during the execution of action a. t The processing efficiency is fed back later. B2 is the weighting coefficient for workpiece surface roughness, and f2(t) is the processing efficiency during the execution of action a. t The surface roughness of the workpiece is fed back later.

[0083] S550. Update the Q-values ​​of the current state and action. During the first execution of the Q-learning algorithm, the Q-values ​​Q(s,a) of the state and action are initialized to initial values; typically, these initial values ​​can be set to 0 or a small random value. Subsequent executions of the Q-learning algorithm continuously update the Q-values ​​of the state and action until they stabilize. The updated Q-values ​​are also used in step S520 to select the current action 'a'. t The basis for this is as follows: The formula for updating the Q value is as follows:

[0084]

[0085] Where, Q′(s t ,a t ) is in state s t Next, execute action a t The Q value, i.e., the current action a t The expected cumulative reward value of this strategy; Q(s) t ,a t ) represents the Q-value before the update; α is the learning rate, representing the weight of the influence of new information on the existing Q-value, with a range of 0 < α ≤ 1; r t+1 To perform action a at the current time step t t The immediate reward obtained afterward; γ is a discount factor, representing the degree of influence of future rewards on the current decision, with a value range of 0≤γ≤1; For the next state S t+1 In the middle, all selectable actions a′ t+1 The action with the highest Q value.

[0086] S560. Determine whether a preset termination condition is met. The termination condition is that the system has converged to the optimal processing parameters (e.g., when the change in Q value is less than a certain threshold, or the reward value stabilizes at a high level within several steps, the system can be judged to have converged to the optimal processing parameters), vibration is minimized, or other predetermined conditions are met (e.g., processing efficiency, surface roughness, etc.). If the termination condition is met, it is considered that the control strategy has basically reached its optimum and no further optimization is needed. Return to step S200 to continue monitoring the processing process. If the termination condition is not met, it is considered that the control strategy still has room for further optimization. Then, for the current state s... t After the update, return to step S510 to adjust the control strategy again until the requirements are met.

[0087] In this embodiment, a method combining vibration prediction model for early prediction and reinforcement learning algorithm for optimizing control strategy is used to suppress vibration. The vibration prediction model can improve the prediction of machine tool vibration, thereby enabling the reinforcement learning algorithm to optimize the control strategy in advance before potentially large vibrations occur, avoiding significant vibrations that could affect machining accuracy. Furthermore, by continuously executing the reinforcement learning algorithm to optimize the control strategy during machining, the control strategy can be continuously improved, ultimately enabling the machine tool to achieve optimal machining performance with minimal vibration.

[0088] Please see Figure 5 , Figure 5 This is a structural block diagram of an embodiment of the machine tool vibration suppression and optimization system of the present invention. The machine tool vibration suppression and optimization system of this embodiment includes a control strategy module 100, a data acquisition module 300, a vibration prediction module 400, and a parameter optimization module 500. The control strategy module 100 is used to generate a control strategy based on machining parameters and generate real-time control signals to the machine tool actuator 200 according to the control strategy, thereby controlling the working parameters of the machine tool actuator 200 in each operating cycle.

[0089] The data acquisition module 300 is used to collect vibration data from key parts of the machine tool actuator 200 during machine tool operation. The data acquisition module 300 may include a vibration acquisition unit, a noise reduction unit, and a feature extraction unit. The vibration acquisition unit is used to collect vibration signals from key parts of the machine tool actuator 200, such as the spindle, cutting tool, and worktable. It may include multiple vibration acquisition devices such as accelerometers, displacement sensors, and strain gauges installed at these key parts. The data collected by these vibration acquisition devices can be transmitted to the noise reduction unit via industrial IoT or similar methods. The noise reduction unit uses filters (such as Kalman filters and bandpass filters) to denoise the collected vibration signals, eliminating high-frequency and low-frequency noise. The feature extraction unit uses methods such as Fourier transform and wavelet transform to extract the frequency domain and time domain features of the vibration signals to obtain vibration data.

[0090] The vibration prediction module 400 is used to predict the vibration results of subsequent operating cycles based on vibration data and the machine tool's operating parameters in the current operating cycle. The vibration prediction module 400 can be trained using a machine learning model or a deep learning model, for example, it can be obtained by executing steps S310 to S330 in the embodiment of the machine tool vibration suppression optimization method described above.

[0091] The parameter optimization module 500 is used to adjust the machine tool operating parameters in the control strategy using a reinforcement learning algorithm when the vibration result predicted by the vibration prediction module does not meet the processing requirements. For example, the machine tool operating parameters in the control strategy can be adjusted by executing steps S510 to S560 in the embodiment of the above-described machine tool vibration suppression optimization method, thereby achieving adaptive control of the CNC machine tool.

[0092] The machine tool vibration suppression optimization system in this embodiment can effectively suppress vibration, significantly improve machining accuracy, reduce equipment wear and failures, and extend machine tool service life through real-time vibration prediction and adaptive control. Furthermore, predictive maintenance can reduce sudden failures and downtime, thereby lowering maintenance costs.

[0093] The above embodiments merely illustrate preferred implementations of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention. Therefore, the protection scope of this invention should be determined by the appended claims.

Claims

1. A method for optimizing machine tool vibration suppression, characterized in that, Includes the following steps: S100: Generate a control strategy based on the machining parameters, and generate real-time control signals based on the control strategy to control the working parameters of the machine tool in each operating cycle; S200: Collect vibration data of key parts of the CNC machine tool during operation; S300: Input the vibration data and the machine tool's current operating parameters into the vibration prediction model to predict subsequent vibration results; S400: Determine whether the vibration prediction results predicted by the vibration prediction model meet the processing requirements. If not, proceed to step S500; otherwise, return to step S200. S500: Based on the vibration prediction results predicted by the vibration prediction model, the reinforcement learning algorithm is used to adjust the machine tool working parameters in the control strategy. The method for adjusting machine tool operating parameters in the control strategy using reinforcement learning algorithms includes the following steps: S510. Obtain the current processing parameters and the vibration prediction results from the vibration prediction model, and represent them as the current state s. t ; S520. Based on the current state, select an action a as the current action a using the ϵ-greedy strategy. t ; S530. Adjust the machine tool's operating parameters in the control strategy to execute the current action. t And while performing the current action a t After the predetermined time, real-time vibration data and processing results are obtained as feedback. S540. Calculate the reward value r based on the feedback results. t+1 ; S550, Update the Q value of the current state and action; S560. Determine whether the preset termination condition is met. If the termination condition is met, return to step S200; otherwise, return to step S510. For the current state s... t After updating, the control strategy is adjusted again until the requirements are met; In step S540, action a is performed. t The reward value after r t+1 The calculation formula is as follows: Among them, a v These are the weighting coefficients for the vibration data values; V ( t ) indicates that the machine tool is performing action a t The values ​​of the vibration data after feedback; i This is an index of the number of parameters used in calculating the reward value. N The total number of parameters used in calculating the reward value. b i For the first i The weighting coefficients of the term parameters, f i ( t ) is the first i Item parameter in the execution of action a t The feedback value is calculated as follows: when the parameter is larger, the weighting coefficient is positive; when the parameter is smaller, the weighting coefficient is negative.

2. The machine tool vibration suppression optimization method as described in claim 1, characterized in that: The key components of the machine tool include the spindle, cutting tool, and worktable of the CNC machine tool; the working parameters include cutting speed, cutting acceleration, feed rate, depth of cut, spindle speed, tool type, and cutting path.

3. The machine tool vibration suppression optimization method as described in claim 1, characterized in that: The method for collecting vibration data of key parts of a machine tool includes the following steps: S210. Vibration acquisition devices are installed at key parts of the machine tool to collect vibration signals from the key parts of the machine tool; S220. Denoise the vibration signal to eliminate high-frequency and low-frequency noise; S230. Extract the frequency domain and time domain features of the vibration signal to obtain vibration data.

4. The machine tool vibration suppression optimization method as described in claim 1, characterized in that: The method for obtaining a vibration prediction model includes the following steps: S310. Determine a machine learning model or deep learning model as the base model for the vibration prediction model; S320. Acquire historical vibration data of the machine tool, determine the normal state data and abnormal state data in the historical vibration data according to the vibration range when the machine tool is working normally, and mark the vibration data of the reverse point, start point and stop point as well as the abnormal state data. S330. The basic model is trained using the labeled vibration data to obtain the vibration prediction model.

5. The machine tool vibration suppression optimization method as described in claim 1, characterized in that: The Long Short-Term Memory (LSTM) network is used as the basic model.

6. The machine tool vibration suppression optimization method according to any one of claims 1 to 5, characterized in that: The reinforcement learning algorithm adopts the Q-learning algorithm, which uses the current working parameter value of the machine tool and vibration data as the state S of the Q-learning algorithm, the adjustment of the working parameters of the machine tool as the action a of the Q-learning algorithm, and the reduction of machine tool vibration and / or the optimization of machine tool processing performance as the reward R of the Q-learning algorithm.

7. The machine tool vibration suppression optimization method as described in claim 6, characterized in that: In step S550, the formula for updating the Q value is as follows: in, In the state s t Next, execute action a t The Q value, i.e., the current action a t The expected cumulative reward value for this strategy; This represents the Q value before the update; α The learning rate represents the weight of the influence of new information on the existing Q value, and its value ranges from 0 to 1. r t+1 To step at the current time t Perform action a t The instant reward obtained afterward; γ The discount factor represents the degree of influence of future rewards on the current decision, and its value ranges from 0 to 1. γ ≤1; For the next state S t+1 In the middle, all selectable actions a′ t+1 The action with the highest Q value.

8. A machine tool vibration suppression optimization system, characterized in that, The machine tool vibration suppression optimization method as described in any one of claims 1 to 7 includes: The control strategy module is used to generate control strategies based on machining parameters and generate real-time control signals to the machine tool actuators based on the control strategies, thereby controlling the working parameters of the machine tool actuators in each operating cycle. The data acquisition module is used to collect vibration data of key parts of the machine tool during operation. The vibration prediction module is trained using a machine learning model or a deep learning model and is used to predict the vibration results of subsequent operating cycles based on vibration data and the machine tool's operating parameters in the current operating cycle. as well as The parameter optimization module is used to adjust the machine tool operating parameters in the control strategy using a reinforcement learning algorithm when the vibration results predicted by the vibration prediction module do not meet the processing requirements.

Images