A method for monitoring the machining state of a numerically controlled machine tool

Abstract

The present application discloses a method for monitoring the machining state of a numerically controlled machine tool, belonging to the technical field of numerically controlled machine tool monitoring, and is used to monitor the vibration state of the milling cutter of the machine tool during the milling process in real time, that is, to predict the vibration impact on the workpiece. The monitoring method includes the following steps: S10, dividing the complete milling process of the milling cutter along the time line into an entry stage, a stable stage, and an exit stage, and respectively obtaining the auto-power spectral density values of the milling cutter in the entry stage, the stable stage, and the exit stage; S20, processing the auto-power spectral density values of the entry stage and the exit stage in a compensatory manner to make them maximally close to the auto-power spectral density value of the stable stage, respectively obtaining a first processed value and a third processed value, calculating and outputting the average value of the auto-power spectral density value of the stable stage, the first processed value, and the third processed value to obtain the vibration state information of the milling cutter of the machine tool. The method for monitoring the machining state of the numerically controlled machine tool provided by the present application obtains accurate and complete vibration state information for monitoring the milling cutter of the machine tool.

Description

Technical Field

[0001] The present invention relates to the technical field of numerical control machine tool monitoring, and particularly to a method for monitoring the machining state of a numerical control machine tool, which is used to monitor the vibration state of the milling cutter of the machine tool during the milling process in real time, that is, to predict the vibration impact on the workpiece. Background Art

[0002] In advanced manufacturing, numerical control machine tools, as core equipment, play an indispensable and important role. The stability during the milling process of the milling cutter is crucial for the machining quality. Especially during high-speed milling at a speed of 500 m / min, it has important theoretical and practical guiding significance for the popularization and application of high-speed milling machining.

[0003] In the prior art, there are many methods for monitoring the machining state of numerical control machine tools. For example, the method and device for monitoring the machining state of a numerical control machine tool with the application number CN116880356A can only roughly monitor the parameters of key parts of the machine tool based on sensors, such as vibration parameters, noise parameters, and temperature parameters, etc., and then use CAE tools to establish a CAE model for simulation analysis. However, there are multiple working stages during the milling process of the milling cutter, including the cutting-in stage, the stable stage, and the cutting-out stage. Among them, the stable stage has the longest duration. There is a problem of inaccurate monitoring data when directly monitoring vibration parameters through sensors. If there are large deviations in the pre-monitoring data used as the data source, no matter how much adjustment and optimization are carried out later, it is difficult to reflect the actual vibration state of the milling cutter.

[0004] For example, the method for monitoring faults during the cutting-in process of a milling cutter based on vibration and acoustic emission sensors with the application number CN202311437936.6 collects sensing data containing tool state, machining process, fault form, etc. through triaxial accelerometers and acoustic emission sensors installed on the numerical control machine tool, monitors the cutting-in state of the milling cutter, and uses a CNN neural network to improve the training speed and accuracy of the tool fault recognition model to prevent production risks that may be caused by continuing to machine after the milling cutter is damaged. It only monitors the cutting-in process of the milling cutter, and there is a problem of incomplete monitoring data, which cannot reflect the complete vibration state of the milling cutter, and thus it is also difficult to analyze the stability of the milling process of the milling cutter. Summary of the Invention

[0005] The embodiments of the present application provide a method for monitoring the machining state of a numerical control machine tool, which can overcome the defects of inaccurate monitoring data and incomplete monitoring data in the prior art, and can thus obtain accurate and complete vibration state information of the milling cutter of the machine tool, so as to timely discover and adjust problems in the machining process of the milling cutter, improve the machining accuracy of the workpiece, and help determine the optimal cutting parameters of the milling cutter to achieve an efficient and stable cutting process.

[0006] An embodiment of the present application provides a method for monitoring the machining state of a numerical control machine tool, which is used to monitor the vibration state of the milling cutter of the machine tool during the milling process in real time, that is, to predict the vibration impact on the workpiece, including the following steps:

[0007] S10, divide the complete milling process of the milling cutter into three stages along the time line, namely the cutting-in stage, the stable stage, and the cutting-out stage, and respectively obtain the auto-power spectral density values of the milling cutter in the cutting-in stage, the stable stage, and the cutting-out stage;

[0008] S20, process the auto-power spectral density values of the cutting-in stage and the cutting-out stage in a compensatory manner to make them maximally close to the auto-power spectral density value of the stable stage, respectively obtain the first processed value and the third processed value, calculate and output the average of the auto-power spectral density value of the stable stage, the first processed value, and the third processed value, and obtain the vibration state information of the milling cutter of the machine tool.

[0009] In a possible implementation manner, in step S10, the auto-power spectral density values of the milling cutter in the cutting-in stage, the stable stage, and the cutting-out stage are respectively obtained through the vibration model of the milling cutter of the machine tool, and the establishment method of the vibration model of the milling cutter of the machine tool includes:

[0010] Taking the predetermined milling direction as the reference, divide the milling area into n segments, and respectively extract the characteristic parameters of the milling cutter in the feed speed direction, the milling width direction, and the milling depth direction in each segment to form a sub-paragraph milling area database, where n is greater than or equal to 50;

[0011] In the sub-paragraph milling area database, use the counts of the n segments as 1, 2, 3... n as the horizontal columns, and use the feed speed parameter, the milling width parameter, and the milling depth parameter as the parallel vertical columns to form a milling data matrix;

[0012] Respectively compare the feed speed parameter, the milling width parameter, and the milling depth parameter corresponding to each horizontal column, and arrange them in ascending or descending order according to the numerical size to form the first data set, the second data set, and the third data set respectively;

[0013] Respectively remove the largest and smallest several numbers in the first data set, the second data set, and the third data set, and then remove the entire columns corresponding to the sub-paragraph milling area database, including the repeated and non-repeated columns in the first data set, the second data set, and the third data set, to form a screened milling area database, including a cutting-in stage milling database, a stable stage milling database, and a cutting-out stage milling database, where the number of horizontal columns in the screened milling area database is greater than or equal to 30, and when the number of horizontal columns in the sub-paragraph milling area database is less than 30, reduce the value of the several numbers or increase the value of n, and re-arrange and screen to make the number of horizontal columns greater than or equal to 30;

[0014] Extract the characteristic parameters of the milling database in the cutting-in stage, the steady stage, and the cutting-out stage respectively, and use the Gaussian process model to construct a machine tool milling cutter vibration model with three stages.

[0015] In a possible implementation, in step S20, the auto-power spectral density values in the cutting-in stage and the cutting-out stage are processed in a compensated manner to maximize their approximation to the auto-power spectral density value in the steady stage. Specifically: establish a compensation model based on the output of the neural network to minimize the difference between the auto-power spectral density values in the cutting-in stage and the cutting-out stage and the auto-power spectral density value in the steady stage, where the neural network is a multi-layer perceptron (MLP) or a radial basis function network (RBFN).

[0016] In a possible implementation, establishing the compensation model based on the output of the neural network specifically means establishing the compensation model through the output of the multi-layer perceptron (MLP);

[0017] The multi-layer perceptron (MLP) includes three input layers, two hidden layers, and one output layer. Among them, the three input layers respectively correspond to the feed speed parameter, the milling width parameter, and the milling depth parameter in the milling database in the cutting-in stage. Input the three input layers into the fully connected network of the two hidden layers, then activate through the activation function ReLU and output through the output layer to establish the compensation model in the cutting-in stage, and then calculate the compensation coefficient in the cutting-in stage. Then, establish the compensation model in the cutting-out stage in the same way, calculate the compensation coefficient in the cutting-out stage, and then adjust the cutting parameters based on the vibration conditions monitored during the actual cutting process to continuously dynamically update and optimize the compensation models in the cutting-in stage and the cutting-out stage, that is, update and optimize the compensation coefficients in the cutting-in stage and the cutting-out stage.

[0018] In a possible implementation, during the process of establishing the compensation model based on the output of the neural network, first extract the frequency characteristics of the feed speed parameter, the milling width parameter, and the milling depth parameter in the milling database in the cutting-in stage and the feed speed parameter, the milling width parameter, and the milling depth parameter in the milling database in the cutting-out stage to determine the frequency range to be compensated;

[0019] Design an IIR filter according to the extracted frequency characteristic results to determine the frequency range, order, and bandwidth of the filter, and perform filtering processing on the milling database in the cutting-in stage and the milling database in the cutting-out stage respectively, and then input them into the fully connected network.

[0020] In a possible implementation, verify the compensation effects of the compensation coefficients in the cutting-in stage and the cutting-out stage respectively through the correlation coefficient method. If the compensation effect exceeds the predetermined threshold, cycle through increasing the additional filtering stage of the filter, performing band-stop filtering, and adaptive filtering in turn until the compensation effect is within the predetermined threshold range.

[0021] Beneficial effects: Compared with the prior art, the method for monitoring the machining state of a numerically controlled machine tool provided by this application obtains the auto-power spectral density values of the milling cutter in the cutting-in stage, the stable stage, and the cutting-out stage respectively, and at the same time processes the auto-power spectral density values in the cutting-in stage and the cutting-out stage in a compensated manner to make them maximize and approach the auto-power spectral density value in the stable stage. Then, the average value of the three processed values is obtained to obtain the vibration state information of the machine tool milling cutter. Thus, the vibration information of the milling cutter in different stages during the cutting process can be fully considered, and further, the integrity and accuracy of the monitoring data can be ensured. Thereby, various problems occurring during the machining process of the milling cutter can be discovered and adjusted in a timely manner, the machining accuracy and machining stability of the workpiece can be improved, the wear and damage conditions of the milling cutter can be monitored and the milling cutter can be replaced in a timely manner, which helps to determine the optimal cutting parameters of the milling cutter, realize an efficient and stable cutting process, and helps to optimize the machine tool structure, improve the rigidity and stability of the machine tool, and thus enhance the overall performance of the machine tool;

[0022] Among them, by dividing the milling area into n segments and respectively extracting the characteristic parameters of the milling cutter in the feed speed direction, the milling width direction, and the milling depth direction in each segment, and then stripping and screening the data set through a specific method to form a cutting-in stage milling database, a stable stage milling database, and a cutting-out stage milling database, and finally extracting characteristic parameters based on these databases to construct a machine tool milling cutter vibration model with three stages, the unreliable factors in the sub-segment milling area database can be effectively stripped, ensuring the accuracy of the machine tool milling cutter vibration model;

[0023] Among them, by combining the neural network algorithm and the compensation model to calculate the compensation coefficients of the milling cutter in the cutting-in stage and the cutting-out stage, the powerful non-linear mapping ability and self-learning ability of the neural network, as well as the accuracy of the compensation model, can be fully utilized to more accurately estimate the vibration compensation amount in the cutting-in stage and the vibration compensation amount in the cutting-out stage of the milling cutter. Thus, the prediction accuracy of the milling cutter vibration state can be ensured, and further, the integrity and accuracy of the numerically controlled machine tool machining state monitoring data can be ensured.

[0024] These and other objects, features, and advantages of the present invention are fully embodied through the following detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 Shows a schematic diagram of the principle of the method for monitoring the machining state of the numerically controlled machine tool of this application.

[0026] Figure 2 Shows an example schematic diagram of screening the milling area database of this application. DETAILED DESCRIPTION OF THE INVENTION

[0027] The following description is used to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments in the following description are only examples, and other obvious variations can be conceived by those skilled in the art. The basic principles of the present invention defined in the following description can be applied to other implementation schemes, variant schemes, improvement schemes, equivalent schemes, and other technical schemes without departing from the spirit and scope of the present invention.

[0028] Those skilled in the art should understand that in the disclosure of the specification, the orientation or positional relationship indicated by terms such as "longitudinal", "transverse", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc. is based on the orientation or positional relationship shown in the drawings. It is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation. Therefore, the above terms should not be construed as limiting the present invention.

[0029] It can be understood that the term "a" should be understood as "at least one" or "one or more". That is, in one embodiment, the number of an element can be one, while in other embodiments, the number of the element can be multiple. The term "a" should not be construed as limiting the quantity.

[0030] During the experimental study on the stability of the machine tool milling cutter vibration system, it was found that the milling process of the machine tool milling cutter can be roughly divided into three stages, namely, the cutting-in stage when the milling cutter just contacts the workpiece, the stable stage after the contact is completed, and the cutting-out stage when the milling cutter moves closer to the sensor position. Among them, the change in the auto-power spectral density value of the milling cutter vibration response corresponds to these three stages respectively. That is, in the cutting-in stage, due to the cutting-in impact, the vibration is significant and the auto-power spectral density value is the largest, and then it rapidly decreases. In the stable stage, the spectral value fluctuates insignificantly. In the cutting-out stage, due to the cutting-out impact, the vibration response gradually strengthens again until the milling cutter completely leaves the workpiece and stops machining, and the vibration of the workpiece will reach an extreme value. Therefore, there are significant differences in the auto-power spectral density values corresponding to different stages. The maximum power spectral density value can be several times the minimum power spectral density value, which also indicates that there are significant differences in the impact on the vibration of the workpiece during the milling process, and the maximum vibration energy value is several times the minimum vibration energy value.

[0031] Therefore, in the monitoring of the machining state of CNC machine tools in the prior art, there is an inaccuracy problem in directly monitoring the vibration information of the milling cutter through sensors (such as displacement sensors, velocity sensors, acceleration sensors, piezoelectric sensors). For example, different monitoring times may monitor data in different stages. Also, even through multiple monitoring and averaging methods, it is still difficult to make up for this deviation because there are large irrational deviations in the data in the cutting-in stage and the cutting-out stage. Even by substituting more data in the stable stages to balance this irrational deviation, it still affects the accuracy of the data. In addition, only monitoring the state of the milling cutter in a certain stage, such as the state of the cutting-in stage disclosed in the comparative document, or even simultaneously monitoring the states of the cutting-in stage, the stable stage, and the cutting-out stage, will still result in incomplete and inaccurate monitoring data because it is difficult to accurately and clearly divide these three stages, and the states of these three stages are not fused. Moreover, in the actual operation process, simultaneously and real-time monitoring the states of the three stages will seriously increase the monitoring task, with high monitoring costs and high maintenance costs.

[0032] Therefore, how to fully and effectively fuse the three stages in the milling process of the milling cutter to ensure the accuracy and integrity of monitoring the vibration state during the milling process of the milling cutter, that is, the accuracy and integrity of predicting the impact on the vibration of the workpiece, is a major problem in the current field of advanced manufacturing CNC machine tools.

[0033] Reference Figure 1 , an embodiment of the present application provides a method for monitoring the machining state of a CNC machine tool, which is used to monitor the vibration state of the milling cutter of the machine tool during the milling process in real time, that is, to predict the impact on the vibration of the workpiece, and includes the following steps:

[0034] S10, divide the complete milling process of the milling cutter into three stages along the time line, namely the cutting-in stage, the stable stage, and the cutting-out stage, and respectively obtain the auto-power spectral density values of the milling cutter in the cutting-in stage, the stable stage, and the cutting-out stage;

[0035] S20, process the auto-power spectral density values of the cutting-in stage and the cutting-out stage in a compensatory manner to make them maximally close to the auto-power spectral density value of the stable stage, respectively obtain the first processed value and the third processed value, calculate and output the average of the auto-power spectral density value of the stable stage, the first processed value, and the third processed value, and obtain the vibration state information of the milling cutter of the machine tool.

[0036] Therefore, the method for monitoring the machining state of the numerical control machine tool provided by this application compensates the auto power spectral density values in the cutting-in stage and the cutting-out stage, making it maximize the approximation to the auto power spectral density value in the stable stage. So that during the actual monitoring process, the vibration information of the milling cutter in any stage is basically accurate vibration information, which is convenient for monitoring and has high monitoring accuracy. At the same time, in the monitoring method provided by this application, the average value of the first processed value, the auto power spectral density value in the stable stage, and the third processed value is further calculated, which can maximize the accuracy and integrity of the vibration state information of the machine tool milling cutter, and can easily and accurately analyze the vibration influence of the machine tool milling cutter on the workpiece. Furthermore, various problems occurring during the machining process of the milling cutter can be timely discovered and adjusted, improving the machining accuracy and machining stability of the workpiece. At the same time, the wear and damage conditions of the milling cutter can also be monitored to replace the milling cutter in time, and it helps to determine the optimal cutting parameters of the milling cutter to achieve an efficient and stable cutting process.

[0037] In one embodiment, in step S10, the auto power spectral density values of the milling cutter in the cutting-in stage, the stable stage, and the cutting-out stage are respectively obtained through the machine tool milling cutter vibration model, and the method for establishing the machine tool milling cutter vibration model includes:

[0038] Based on the predetermined milling direction, the milling area is divided into n segments, and the characteristic parameters of the milling cutter in the feed speed direction, the milling width direction, and the milling depth direction in each segment are respectively extracted to form a sub-paragraph milling area database, where n is greater than or equal to 50;

[0039] Then, in the sub-paragraph milling area database, with the counts of the n segments being 1, 2, 3... n as the horizontal columns respectively, and with the feed speed parameter, the milling width parameter, and the milling depth parameter as the parallel vertical columns, a milling data matrix is formed;

[0040] Then, the feed speed parameter, the milling width parameter, and the milling depth parameter corresponding to each horizontal column are respectively compared, and arranged in ascending or descending order according to the numerical size to form a first data set, a second data set, and a third data set respectively;

[0041] Then, remove a certain number of the largest and smallest numbers from the first dataset, the second dataset, and the third dataset respectively. Then, remove the entire column corresponding to the sub-paragraph milling area database, including the repeated and non-repeated columns in the first dataset, the second dataset, and the third dataset. Recombine the remaining data (the combination order is not important because the final average is calculated, and it can be combined by column or in ascending or descending order of a certain row of data) to form a filtered milling area database, including the cutting-in phase milling database, the stable phase milling database, and the cutting-out phase milling database. Among them, the number of horizontal columns in the filtered milling area database is greater than or equal to 30. When the number of horizontal columns in the sub-paragraph milling area database is less than 30, reduce the value of the certain number or increase the value of n, and rearrange and filter again to make the number of horizontal columns greater than or equal to 30 to ensure that there is enough data to support the accuracy of the monitoring results;

[0042] Finally, extract the characteristic parameters from the cutting-in phase milling database, the stable phase milling database, and the cutting-out phase milling database respectively, and use the Gaussian process model to construct a machine tool milling cutter vibration model with three phases. In this way, it is possible to accurately divide the milling area into the cutting-in phase, the stable phase, and the cutting-out phase with sufficient data support. At the same time, it is also possible to filter out the irrational data of the feed speed parameter, the milling width parameter, and the milling depth parameter in the three phases respectively, ensuring the accuracy of the data. Finally, by extracting the characteristic parameters from the cutting-in phase milling database, the stable phase milling database, and the cutting-out phase milling database, and using the Gaussian process model to construct a machine tool milling cutter vibration model with three phases, it is a complete and accurate machine tool milling cutter vibration model that can accurately obtain the auto-power spectral density values of the milling cutter in the cutting-in phase, the stable phase, and the cutting-out phase during the milling process.

[0043] The Gaussian Process (GP) is a type of stochastic process in probability theory and mathematical statistics, which is a combination of a series of random variables obeying the normal distribution within an index set.

[0044] The linear combination of any random variable in the Gaussian process obeys the normal distribution, and each finite-dimensional distribution is a joint normal distribution. Moreover, its probability density function on the continuous index set is the Gaussian measure of all random variables. Therefore, it is regarded as an infinite-dimensional generalized extension of the joint normal distribution. The Gaussian process is completely determined by its mathematical expectation and covariance function and inherits many properties of the normal distribution.

[0045] An example of screening the milling area database is as follows. Based on a predetermined milling direction (such as the cutting y direction), the milling area is divided into n segments, forming n horizontal columns. The feed speed parameters, milling width parameters, and milling depth parameters corresponding to each horizontal column are arranged in ascending or descending order according to the numerical values, respectively forming a first data set, a second data set, and a third data set. Then, the two largest numbers and the two smallest numbers in each data set are removed (the number of the largest and smallest numbers removed can be the same or different), a total of 3 * 4 = 12 numbers. Then, the entire columns corresponding to the 12 numbers just removed are removed. If two of the 12 numbers correspond to the same column at the same time, a total of 11 columns are removed. If six of the 12 numbers respectively correspond to two columns at the same time, a total of 9 columns of data are removed. For example, referring to Figure 2 , in the feed speed parameters, the 3rd, 6th, 10th, and nth are removed as the two largest numbers and the two smallest numbers. In the milling width parameters, the 2nd, 6th, 8th, and 14th are removed as the two largest numbers and the two smallest numbers. In the milling depth parameters, the 2nd, 6th, 8th, and 14th are removed as the two largest numbers and the two smallest numbers. Then, the data in the 2nd, 3rd, 5th, 6th, 8th, 10th, 13th, 14th, and nth columns are all removed. This can ensure the accuracy of the feed speed parameters, milling width parameters, and milling depth parameters. Furthermore, the cutting-in stage, stable stage, and cutting-out stage of the milling area of the milling cutter can be divided as accurately as possible through big data calculation. Thus, the self-power spectral density values of the milling cutter in the cutting-in stage, stable stage, and cutting-out stage can be accurately obtained through the machine tool milling cutter vibration model, providing data support for the accurate determination of the cutting-in stage compensation coefficient and the cutting-out stage compensation coefficient in the subsequent stage.

[0046] In addition, it should be noted that in order to ensure the accuracy of the data, there must be enough basic data. Therefore, it is necessary to ensure that the number of horizontal columns in the screened milling area database is greater than or equal to 30. When the number of horizontal columns in the sub-paragraph milling area database is less than 30, the values of the several numbers can be reduced or the value of n can be increased, and then the arrangement and screening are carried out again to make the number of horizontal columns greater than or equal to 30.

[0047] Among them, in the initially formed sub-paragraph milling area database, where n is greater than or equal to 50, it can be realized by a precision robot or by computer simulation.

[0048] In one embodiment, in step S20, the self-power spectral density values in the cutting-in stage and the cutting-out stage are processed in a compensated manner to maximize their approximation to the self-power spectral density value in the steady stage. Specifically: a compensation model is established based on the output of a neural network to minimize the difference between the self-power spectral density values in the cutting-in stage and the cutting-out stage and the self-power spectral density value in the steady stage. The neural network is a multi-layer perceptron (MLP) or a radial basis function network (RBFN). Furthermore, the powerful non-linear mapping ability of the neural network can capture complex patterns in the data, thereby improving the prediction accuracy of the compensation model, minimizing the difference between the self-power spectral density values in the cutting-in stage and the cutting-out stage and the self-power spectral density value in the steady stage. At the same time, the neural network can also make predictions on unseen data and has good generalization ability. In addition, it can automatically adjust the model parameters according to the changes in the input data, enabling the compensation model to adapt to different milling conditions and reducing the time of manual intervention and debugging through automatic prediction and compensation, with high work efficiency.

[0049] Further preferably, establishing the compensation model based on the output of the neural network specifically means establishing the compensation model through the output of the multi-layer perceptron (MLP);

[0050] The multi-layer perceptron (MLP) includes three input layers, two hidden layers, and one output layer. Among them, the three input layers respectively correspond to the feed speed parameter, the milling width parameter, and the milling depth parameter in the milling database in the cutting-in stage. The three input layers are input into the fully connected network of the two hidden layers, and then activated by the activation function ReLU and output through the output layer to establish the cutting-in stage compensation model. Furthermore, in the actual milling process, various parameters (including the feed speed parameter, the milling width parameter, and the milling depth parameter of the milling cutter and the relevant information of the workpiece) can be directly substituted to calculate the cutting-in stage compensation coefficient. Then, the cutting-out stage compensation model is established in the same way to calculate the cutting-out stage compensation coefficient. Then, the cutting parameters are adjusted based on the vibration conditions monitored during the actual cutting process. The cutting parameters include the feed speed, the milling width, and the milling depth, and the cutting-in stage compensation model and the cutting-out stage compensation model can be continuously updated and optimized dynamically, that is, the cutting-in stage compensation coefficient and the cutting-out stage compensation coefficient are updated and optimized. The activation formula of the activation function is: In this way, during the process of establishing the compensation model, the feed speed, milling width, and milling depth of the milling cutter can be fully considered through a multi-layer perceptron to ensure the accuracy of the compensation coefficients in the cutting-in stage and the cutting-out stage. Of course, this is a generalized compensation model or made based on fixed workpiece information. In the actual working process, the situation of the workpiece needs to be specifically considered. Once the information of the workpiece to be processed is determined, the vibration state information of the machine tool milling cutter can be accurately monitored directly through the machine tool milling cutter vibration model and the compensation model. The information of the workpiece to be processed includes the hardness, toughness, thermal conductivity, and related mechanical property parameters of the workpiece (such as yield strength, elongation, etc.).

[0051] For example, during the determination process of the compensation coefficient in the cutting-in stage, referring to the simplified formula Kv = f(V, W, D, T), where:

[0052] Kv is the compensation coefficient in the cutting-in stage.

[0053] V is the feed speed parameter.

[0054] W is the cutting width parameter.

[0055] D is the cutting depth parameter.

[0056] T is the workpiece information. When mainly considering the toughness of the workpiece, the workpiece toughness parameter ∫σdε can be substituted into the formula. When it is necessary to consider the hardness, toughness, and yield strength of the workpiece simultaneously, the specific proportion of each parameter can be determined based on actual needs, and then the T value can be calculated, such as T = hardness value * 30% + toughness value * 30% + yield strength value * 40%.

[0057] f is a function representing the relationship between the compensation coefficient in the cutting-in stage and the cutting parameters and workpiece information, which is directly obtained from the cutting-in stage compensation model.

[0058] In one embodiment, during the process of establishing the compensation model based on the output of the neural network, first, frequency feature extraction is performed on the feed speed parameter, milling width parameter, and milling depth parameter in the cutting-in stage milling database and the feed speed parameter, milling width parameter, and milling depth parameter in the cutting-out stage milling database to determine the frequency range that needs to be compensated;

[0059] Design an IIR filter according to the extracted frequency feature results to determine the frequency range, order, and bandwidth of the filter. Filter the milling database in the cutting-in stage and the milling database in the cutting-out stage respectively, and then input them into the fully connected network. Thus, the signals of the milling cutter in the cutting-in stage and the cutting-out stage can be processed by this IIR filter specially designed for the feed speed parameter, milling width parameter, and milling depth parameter. After obtaining the compensated signals, input them to the multi-layer perceptron (MLP), which can further improve the accuracy of the compensation coefficients in the cutting-in stage and the cutting-out stage.

[0060] In one embodiment, the compensation effects of the compensation coefficients in the cutting-in stage and the cutting-out stage are verified respectively by the correlation coefficient method. If the compensation effect exceeds the predetermined threshold, cycle through increasing the additional filtering stage of the filter, performing band-stop filtering, and adaptive filtering in sequence until the compensation effect is within the predetermined threshold range. If the ideal effect still cannot be achieved after cycling n times, recalculate and simulate for verification, where n is greater than or equal to 50.

[0061] Those skilled in the art should understand that the embodiments of the present invention described above and shown in the accompanying drawings are only examples and do not limit the present invention. The advantages of the present invention have been fully and effectively realized. The functions and structural principles of the present invention have been shown and described in the embodiments. Without departing from the said principles, the embodiments of the present invention can have any deformation or modification.

Claims

1. A method for monitoring the machining state of a numerically controlled machine tool, which is used to monitor the vibration state of the milling cutter of the machine tool during the milling process in real time, that is, to predict the vibration impact on the workpiece, and is characterized in that Including the following steps: S10: Divide the complete milling process of the milling cutter along the time line into three stages, namely the cutting-in stage, the stable stage, and the cutting-out stage, and respectively obtain the auto-power spectral density values of the milling cutter in the cutting-in stage, the stable stage, and the cutting-out stage; S20: Process the auto-power spectral density values of the cutting-in stage and the cutting-out stage in a compensatory manner to make them maximize the approximation to the auto-power spectral density value of the stable stage, respectively obtain the first processed value and the third processed value, calculate and output the average of the auto-power spectral density value of the stable stage, the first processed value, and the third processed value, and obtain the vibration state information of the machine tool milling cutter; In step S10, the auto-power spectral density values of the milling cutter in the cutting-in stage, the stable stage, and the cutting-out stage are respectively obtained through the vibration model of the machine tool milling cutter, and the establishment method of the vibration model of the machine tool milling cutter includes: Taking the predetermined milling direction as the reference, divide the milling area into n segments, respectively extract the characteristic parameters of the milling cutter in the feed speed direction, the milling width direction, and the milling depth direction in each segment, and form a sub-paragraph milling area database, where n is greater than or equal to 50; In the sub-paragraph milling area database, use the counts of the n segments as 1, 2, 3... n as the horizontal columns, and use the feed speed parameter, the milling width parameter, and the milling depth parameter as the parallel vertical columns to form a milling data matrix; Respectively compare the feed speed parameter, the milling width parameter, and the milling depth parameter corresponding to each horizontal column, and arrange them in ascending or descending order according to the numerical size, respectively forming a first data set, a second data set, and a third data set; Respectively remove the largest and smallest several numbers in the first data set, the second data set, and the third data set, and then remove the entire columns corresponding to the sub-paragraph milling area database, including the repeated and non-repeated columns in the first data set, the second data set, and the third data set, to form a filtered milling area database, including the cutting-in stage milling database, the stable stage milling database, and the cutting-out stage milling database, where the number of horizontal columns in the filtered milling area database is greater than or equal to 30, and when the number of horizontal columns in the sub-paragraph milling area database is less than 30, reduce the value of the several numbers or increase the value of n, and re-arrange and filter to make the number of horizontal columns greater than or equal to 30; Respectively extract the characteristic parameters of the cutting-in stage milling database, the stable stage milling database, and the cutting-out stage milling database, and use the Gaussian process model to construct a vibration model of the machine tool milling cutter with three stages; 2. The method for monitoring the machining state of a numerically controlled machine tool according to claim 1, characterized in that, In step S20, the processing of the auto-power spectral density values of the cutting-in stage and the cutting-out stage in a compensatory manner to make them maximize the approximation to the auto-power spectral density value of the stable stage is specifically: establish a compensation model based on the output of the neural network to minimize the difference between the auto-power spectral density values of the cutting-in stage and the cutting-out stage and the auto-power spectral density value of the stable stage, where the neural network is a multi-layer perceptron (MLP) or a radial basis function network (RBFN); 3. The method for monitoring the machining state of a numerically controlled machine tool according to claim 2, characterized in that, The establishment of the compensation model based on the output of the neural network is specifically to establish a compensation model through the output of the multi-layer perceptron (MLP); The multi-layer perceptron (MLP) includes three input layers, two hidden layers and one output layer. Among them, the three input layers respectively correspond to the feed rate parameter, milling width parameter and milling depth parameter in the milling database during the plunge cutting stage. The three input layers are input into the fully connected network of the two hidden layers, and then activated by the activation function ReLU and output through the output layer to establish a plunge cutting stage compensation model. Furthermore, the plunge cutting stage compensation coefficient is calculated. Then, the compensation model for the outfeed cutting stage is established in the same way, and the outfeed cutting stage compensation coefficient is calculated. Then, the cutting parameters are adjusted based on the vibration conditions monitored during the actual cutting process to continuously and dynamically update and optimize the plunge cutting stage compensation model and the outfeed cutting stage compensation model, that is, to update and optimize the plunge cutting stage compensation coefficient and the outfeed cutting stage compensation coefficient.

4. The method for monitoring the machining state of a numerically controlled machine tool according to claim 3, wherein, During the process of establishing the compensation model based on the output of the neural network, first, frequency feature extraction is performed on the feed rate parameter, milling width parameter and milling depth parameter in the milling database during the plunge cutting stage, as well as the feed rate parameter, milling width parameter and milling depth parameter in the milling database during the outfeed cutting stage to determine the frequency range that needs to be compensated. An IIR filter is designed according to the extracted frequency feature results to determine the frequency range, order and bandwidth of the filter. The milling databases during the plunge cutting stage and the outfeed cutting stage are respectively filtered and then input into the fully connected network.

5. The method for monitoring the machining state of a numerically controlled machine tool according to claim 4, wherein, The compensation effects of the plunge cutting stage compensation coefficient and the outfeed cutting stage compensation coefficient are respectively verified by the correlation coefficient method. If the compensation effect exceeds the predetermined threshold, the additional filtering stage of the filter, band-stop filtering and adaptive filtering are cyclically increased in sequence until the compensation effect is within the predetermined threshold range.

Images