Method, device and equipment for predicting multi-directional stress in milling cutter milling process and storage medium

Abstract

The application provides a multi-directional stress prediction method, device and equipment in a milling cutter milling process and a storage medium. It relates to the field of numerical control machine tool processing digital twin technology. The method comprises: obtaining small sample experimental data based on an orthogonal test method, time-frequency decomposition of the milling force test signal, and extraction of multi-dimensional features of the milling force dynamic characteristics; analyzing the correlation between the processing parameters and the features, screening the key features, establishing a physical mapping model of the process parameters to the key features and solving the cutting coefficients; constructing a time-varying signal prediction model based on a recurrent neural network, predicting the multi-directional dynamic milling force of the milling cutter with the key features in the small sample test data; and based on the cutting coefficient, designing an adaptive filter to post-process and optimize the predicted signal and inverse normalize it, and output the final prediction value. Based on small sample data, the application can accurately predict the multi-directional dynamic stress of the milling cutter only with the process parameters, and effectively improve the virtual-real mapping and dynamic optimization capability of the processing process.

Description

Technical Field

[0001] This application relates to the field of digital twin technology for CNC machine tool processing, and in particular to a method, device, equipment and storage medium for predicting multi-directional forces during milling. Background Technology

[0002] Milling, as a fundamental and widely used machining method, relies heavily on the multi-axis linkage and control capabilities of CNC machine tools to achieve high efficiency and precision. Its machining quality and efficiency depend on key process parameters such as cutter selection, spindle speed, feed rate, and depth of cut. Different machining requirements impose different specifications on cutter selection, spindle speed, feed rate, and depth of cut. While traditional materials often allow for process parameters based on tool cutting parameter tables, new materials and special machining requirements often bring various process quality and safety issues. Therefore, digital twin technology based on machine tool system simulation modeling provides a new approach for optimizing milling processes under special working conditions. However, since the real-time accuracy of milling forces directly affects the reliability of the machine tool simulation model, current methods still have significant limitations in accurately guiding process optimization. Therefore, how to achieve high-precision prediction of multi-directional forces on the milling cutter under unknown working conditions based solely on process parameters, using only small sample experimental data, and how to provide high-precision instantaneous force data input for machine tool simulation modeling and digital twin systems, thereby enhancing the reliability of virtual-real interaction and supporting process optimization and safety control, are urgent technical directions that need to be addressed.

[0003] Existing research primarily focuses on enhancing the mechanical stability of milling systems (such as optimizing machine tool structure, improving fixture design, and applying vibration damping devices) to suppress vibration, improve machining accuracy, and enhance surface quality. However, this approach relies heavily on trial and error experiments, resulting in high costs, limited applicability, and a lack of theoretical support for related process optimization. Therefore, there is an urgent need for a predictive method that directly correlates process parameters with dynamic milling forces. This method should accurately predict multi-directional milling forces under unknown working conditions based on only a small amount of experimental data, providing high-fidelity input data for digital twin systems and simulation models. This would significantly improve the realism of their simulations and mitigate machining risks from a mechanistic perspective. Summary of the Invention

[0004] This application provides a method, apparatus, device, and storage medium for predicting multi-directional forces during milling. By combining the knowledge obtained from an accurate solver with the efficiency of metaheuristic algorithms, it enables high-precision fitting and prediction of multi-directional dynamic forces on the milling cutter.

[0005] In a first aspect, this application provides a method for predicting multi-directional forces during milling, including:

[0006] Based on the orthogonal test method, the milling force signal of a small sample is obtained by using the rotational speed, feed rate, depth of cut and milling material as test factors;

[0007] The milling force signal is subjected to time-frequency decomposition processing using empirical mode decomposition to extract multi-dimensional features characterizing the dynamic properties of the milling force. These multi-dimensional features include dominant frequency features, peak force features, total frequency energy features, and dominant frequency amplitude.

[0008] The correlation between the engineering parameters corresponding to the experimental factors and the multi-dimensional features is calculated. The correlation is quantified by the correlation coefficient, and features with a correlation coefficient greater than a set threshold are selected as key features.

[0009] A physical mapping model is constructed to map process parameters to the key features, and the radial cutting coefficient and tangential cutting coefficient are solved through the physical mapping model.

[0010] Based on the aforementioned key features, the dynamic signal of the multi-directional force on the milling cutter is output as the predicted milling force through a time-varying signal prediction model. The time-varying signal prediction model is a time-series model based on a recurrent neural network.

[0011] Based on the oscillation boundary of the constraint force values ​​of the radial cutting coefficient and the tangential cutting coefficient, the predicted milling force is adaptively filtered using a finite impulse response filter, and then the filtered predicted milling force is inversely normalized to finally output the multi-directional force prediction value.

[0012] In one possible design, the formula for calculating the correlation coefficient is:

[0013] (1);

[0014] In the formula, r Represents the correlation coefficient. X i Indicates the engineering parameter value. Y i Represents eigenvalues; i Index representing the value of engineering parameter. k Indicates the total number of engineering parameters. This represents the average value of engineering parameters. This represents the average value of the eigenvalues.

[0015] In one possible design, the milling force signal is subjected to time-frequency decomposition using empirical mode decomposition to extract multi-dimensional features characterizing the dynamic properties of the milling force, including:

[0016] The main frequency characteristics are calculated using the following formula:

[0017] (2);

[0018] In the formula, f dom Main frequency characteristics; c For frequency coefficients; Z Number of teeth; This refers to the vibration offset related to the rotational speed. n Current rotational speed; a p For depth of cut;

[0019] Peak force characteristics are calculated using the following formula:

[0020] (3);

[0021] In the formula, F peak Characterized by peak force. K c The material-related cutting coefficient; α The material shear strain rate; β For tool-chip friction; f For feed rate;

[0022] The total frequency energy characteristic is calculated using the following formula:

[0023] (4);

[0024] In the formula, E total The total frequency energy characteristic; C is the energy conversion parameter; C is the total frequency energy coefficient.

[0025] The main frequency amplitude is calculated using the following formula:

[0026] (5);

[0027] In the formula, Amplitude minus cutting coefficient; Main frequency amplitude characteristics; Z for Number of teeth ; It is a non-linear exponent.

[0028] In one possible design, the physical mapping model includes radial and tangential micro-element cutting force models of the milling cutter, respectively represented as:

[0029] (6);

[0030] (7);

[0031] In the formula, F t For tangential force, F r Radial force, The tangential cutting coefficient is... h ( θ () represents the instantaneous, undeformed chip thickness. dz The height of the axial micro-element, The first cutting edge force coefficient, The radial cutting coefficient is... This is the second cutting edge force coefficient.

[0032] In one possible design, the time-varying signal prediction model employs a dynamic learning rate adjustment strategy during training, and the learning rate calculation formula satisfies:

[0033] (8);

[0034] In the formula, is the current learning rate; T is the initial learning rate; P is the decaying gradient; h is the current iteration number.

[0035] In one possible design, the time-varying signal prediction model is an LSTM network, which includes an input layer, three hidden layers, and an output layer. The input layer receives a multi-dimensional feature vector obtained through empirical mode decomposition. The hidden layers are configured with memory cell vectors to capture the long-term dependencies of the milling force signal. The output layer uses a linear activation function to generate a dynamic signal sequence of multi-directional forces on the milling cutter. The LSTM network is trained using an elastic backpropagation optimization algorithm, and the initial weights of each feature are set according to their correlation coefficient with engineering parameters.

[0036] In one possible design, the oscillation range of the finite impulse response filter is dynamically adjusted based on the solved radial and tangential cutting coefficients.

[0037] Secondly, this application provides a device for predicting multi-directional forces during milling with a milling cutter, the device comprising:

[0038] The signal acquisition module is configured to acquire a small sample of milling force signals based on the orthogonal experimental method, using rotational speed, feed rate, depth of cut, and milling material as experimental factors;

[0039] The feature calculation module is configured to perform time-frequency decomposition processing on the milling force signal using empirical mode decomposition, and extract multi-dimensional features characterizing the dynamic characteristics of the milling force. The multi-dimensional features include the main frequency feature, peak force feature, total frequency energy feature, and main frequency amplitude.

[0040] The feature filtering module is configured to calculate the correlation between the engineering parameters corresponding to the experimental factors and the multi-dimensional features. The correlation is quantified by the correlation coefficient, and features with a correlation coefficient greater than a set threshold are selected as key features.

[0041] The mapping model construction module is configured to construct a physical mapping model from process parameters to the key features, and solve for the radial cutting coefficient and tangential cutting coefficient through the physical mapping model;

[0042] The milling prediction module is configured to output the dynamic signal of the multi-directional force on the milling cutter as the predicted milling force based on the key features and through a time-varying signal prediction model. The time-varying signal prediction model is a time-series model based on a recurrent neural network.

[0043] The adaptive filtering module is configured to adaptively filter the predicted milling force using a finite impulse response filter based on the oscillation boundary of the constraint force values ​​of the radial cutting coefficient and the tangential cutting coefficient, and then perform inverse normalization processing on the filtered predicted milling force to finally output the multi-directional force prediction value.

[0044] Thirdly, embodiments of this application provide an electronic device, including: at least one processor and a memory; the memory stores computer execution instructions; the at least one processor executes the computer execution instructions stored in the memory, causing the at least one processor to perform the multi-directional force prediction method during milling as described in the first aspect and various possible designs of the first aspect.

[0045] Fourthly, embodiments of this application provide a computer-readable storage medium storing computer-executable instructions. When a processor executes the computer-executable instructions, it implements the method for predicting multi-directional forces during milling as described in the first aspect and various possible designs of the first aspect.

[0046] Fifthly, embodiments of this application provide a computer program product, including a computer program that, when executed by a processor, implements the method for predicting multi-directional forces during milling as described in the first aspect and various possible designs of the first aspect.

[0047] The method, apparatus, equipment, and storage medium for predicting multi-directional forces during milling provided in this application have at least the following beneficial effects:

[0048] This application can accurately predict the multi-directional dynamic force of the milling cutter based on small sample data and only using process parameters. The obtained instantaneous force data can not only provide reliable input for machine tool simulation modeling, but also provide key data support for the construction of digital twin systems, thereby improving the virtual-real mapping and dynamic optimization capabilities of the machining process. Attached Figure Description

[0049] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0050] Figure 1 The flowchart of a method for predicting multi-directional forces during milling provided in this application embodiment. Figure 1 ;

[0051] Figure 2 The flowchart of a method for predicting multi-directional forces during milling provided in this application embodiment. Figure 2 ;

[0052] Figure 3 The key feature correlation coefficient matrix diagram provided for the embodiments of this application;

[0053] Figure 4 This is a comparison chart of the original signal and the EMD decomposition residual provided in the embodiments of this application;

[0054] Figure 5 This is a comparison diagram of the predicted signal and the actual signal provided in the embodiments of this application;

[0055] Figure 6 A flowchart illustrating the overall process of the training model provided in this application embodiment;

[0056] Figure 7 This is a structural diagram of the multi-directional force prediction device provided in the embodiments of this application during the milling process.

[0057] The accompanying drawings illustrate specific embodiments of this application, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concept of this application to those skilled in the art through reference to particular embodiments. Detailed Implementation

[0058] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0059] The collection, storage, use, processing, transmission, provision, and disclosure of financial data or user data involved in the technical solution of this application all comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0060] It should be noted that in the embodiments of this application, certain software, components, models and other existing solutions in the industry may be mentioned. These should be regarded as exemplary and are only intended to illustrate the feasibility of implementing the technical solution of this application. However, it does not mean that the applicant has used or necessarily used the solution.

[0061] The technical solution of this application and how the technical solution of this application solves the above-mentioned technical problems are described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of this application will now be described with reference to the accompanying drawings.

[0062] This application provides a method for predicting multi-directional forces during milling, aiming to quantitatively characterize the tangential and radial force coefficients during milling through signal analysis and modeling techniques, reveal the influence mechanism of process parameters on the cutting coefficient, and establish a predictive model for multi-directional forces on the milling cutter. Figure 1 The diagram shows a flowchart of a method for predicting multi-directional forces during milling, as provided in an embodiment of this application. Figure 1 The implementation process of the multi-directional force prediction method during the milling process of this milling cutter is as follows:

[0063] First, based on the orthogonal experimental design, experimental data were sampled using rotational speed, feed rate, depth of cut, and milling material as factors. Time-frequency decomposition was performed on the collected milling process data to extract multi-dimensional features characterizing the milling force signal. Then, the quantitative characterization of the chip coefficient was analyzed based on calibration data. Subsequently, the correlation strength between process parameters (such as rotational speed, feed rate, depth of cut, and material) and multi-dimensional features was quantified by calculating the correlation coefficient matrix, thus clarifying the influencing mechanism.

[0064] Based on the strongly correlated features and process parameters identified in the above analysis, a time-varying mapping model between the features and the forces acting on the milling cutter in each direction is constructed. The model building process includes: normalizing the force data in the training set to eliminate the influence of the data dimensions; establishing a time-varying fitting relationship between the strongly correlated features and the forces acting on the milling cutter; and introducing a cutting force coefficient to characterize the material properties in order to enhance the model's generalization ability to unknown materials, while also strengthening the model's adaptive learning ability for unknown process parameters and unknown materials.

[0065] Finally, post-processing optimization is performed based on the established engineering parameter-force characteristic mapping relationship: according to the force value oscillation boundary constrained by the cutting coefficient, an adaptive FIR filter is designed to filter the prediction results, and the final multi-directional force prediction value of the milling cutter is output through inverse normalization processing, realizing the closed-loop mapping from process parameters to cutting response.

[0066] like Figure 2The diagram shows a flowchart of a method for predicting multi-directional forces during milling, as provided in an embodiment of this application. Figure 2 The following will combine Figure 2 The specific implementation steps shown detail the implementation process of the multi-directional force prediction method during milling. This multi-directional force prediction method during milling includes the following steps S10-S60.

[0067] S10: Based on the orthogonal test method, the milling force signal of a small sample is obtained by using the rotational speed, feed rate, depth of cut and milling material as test factors.

[0068] In this embodiment, when conducting the orthogonal test method, a certain brand of ordinary vertical three-axis milling machine was used, the milling cutter was a four-tooth flat end mill, and the cutting material was aluminum. In this embodiment, only the milling force prediction of unknown process parameters under the same material was tested.

[0069] S20: The milling force signal is processed by empirical mode decomposition for time and frequency decomposition, and multi-dimensional features characterizing the dynamic characteristics of the milling force are extracted. The multi-dimensional features include the main frequency feature, peak force feature, total frequency energy feature and main frequency amplitude.

[0070] Given the high dimension of the input features, this embodiment will explain the calculation method of the relevant feature projection based on the milling force signal obtained in S10 as follows.

[0071] The formula for calculating the main frequency characteristic is:

[0072] (2);

[0073] In the formula, f dom Main frequency characteristics; c For frequency coefficients; Z Number of teeth; This refers to the vibration offset related to the rotational speed. n The current rotational speed; a p For depth of cut;

[0074] The formula for calculating peak force characteristics is:

[0075] (3);

[0076] In the formula, F peak Characterized by peak force. K c The material-related cutting coefficient; α The material shear strain rate; β For tool-chip friction; f For feed rate;

[0077] The formula for calculating the total frequency energy characteristic is:

[0078] (4);

[0079] In the formula, E total The total frequency energy characteristic; C is the energy conversion parameter; C is the total frequency energy coefficient.

[0080] main frequency amplitude The physical basis model for characterizing the dynamic force intensity at the frequency through which the tool passes is:

[0081] (5);

[0082] In the formula, Amplitude minus cutting coefficient; Main frequency amplitude characteristics; Z Number of teeth; It is a non-linear exponent.

[0083] S30: Calculate the correlation between the engineering parameters corresponding to the experimental factors and the multi-dimensional features. The correlation is quantified by the correlation coefficient, and features with a correlation coefficient greater than a set threshold are selected as key features.

[0084] There is a certain functional relationship between milling process parameters and force signal characteristics (multi-dimensional characteristics). Taking the force signals of a certain direction at 1200 rpm, 2400 rpm and 4800 rpm of the equipment as examples, 5 experimental samples were taken for each (sample sampling frequency is 8000 Hz). After Fourier transform, the main frequency points (frequency points with a total energy ratio of more than 90%) were extracted, as shown in Table 1.

[0085] Table 1. Main frequency distribution of each sample

[0086]

[0087] As shown in Table 1, under the condition of constant rotational speed, all samples retain their inherent frequency signals. With increasing milling depth, the inherent frequency remains constant, but the proportion of high-frequency signals gradually increases. Therefore, solving the mapping relationship between engineering parameters and data features is feasible. Thus, the signals are decomposed using EMD, and all signal features (primary and secondary frequencies, primary and secondary frequency amplitudes, primary and secondary frequency energy ratios, and the average values ​​of the upper and lower envelopes, totaling 111 features) are extracted from all decomposed intrinsic models and residuals. Based on this, the Pearson correlation coefficient between the signal and the features is calculated, as shown in Formula 1:

[0088] (1);

[0089] In the formula, rRepresents the correlation coefficient. X i Indicates the engineering parameter value. Y i Represents eigenvalues; i Index representing the value of engineering parameter. k Indicates the total number of engineering parameters. This represents the average value of engineering parameters. This represents the average of the eigenvalues.

[0090] Based on this example, solving Equation 1 finally yields the correlation coefficient matrix, and the features with correlation coefficients greater than 0.6 with engineering parameters are visualized as follows: Figure 3 As shown.

[0091] S40: Construct a physical mapping model from process parameters to key features, and solve for the radial cutting coefficient and tangential cutting coefficient through the physical mapping model.

[0092] In this embodiment, the radial and tangential micro-element cutting force models are established based on the training samples as follows:

[0093] (6);

[0094] (7);

[0095] In the formula, F t For tangential force, F r Radial force, The tangential cutting coefficient is... h ( θ () represents the instantaneous, undeformed chip thickness. dz The height of the axial micro-element, The first cutting edge force coefficient, The radial cutting coefficient is... This is the second cutting edge force coefficient.

[0096] Based on obtaining sufficient cutting force coefficients, an appropriate regression model is constructed using these coefficients as input. Combined with the quantitative parameters of the material, this model is used to predict the cutting force coefficients of unknown materials under working conditions.

[0097] S50: Based on key features, the dynamic signal of the multi-directional force on the milling cutter is output as the predicted milling force through the time-varying signal prediction model. The time-varying signal prediction model is a time series model based on a recurrent neural network.

[0098] After extracting key features through feature engineering, the time-frequency characteristics of the force signal to be fitted should be analyzed first, including its periodicity and time-varying characteristics. At the same time, targeted preprocessing should be carried out according to the data distribution characteristics, and a suitable regression model should be constructed as a time-varying signal prediction model in combination with the actual application requirements.

[0099] In this example, by comparing the original signal with the EMD decomposition residual signal, it can be seen that (e.g.) Figure 4 As shown in the figure, the original signal contains complex dynamic fluctuations and also implies a potential overall trend of change; while the residual signal exhibits significant monotonicity. This indicates that the signal possesses a continuously stable time-varying characteristic, making the LSTM neural network architecture suitable for modeling and analysis.

[0100] To address the issue of insufficient model generalization caused by insufficient sample size and excessive parameter dimensionality, this example implements a dual data augmentation strategy to expand the training set and improve model robustness. First, the force signal is linearly scaled within the range of 90%-110% to expand the coverage of the parameter space. Then, a random translation with a maximum offset of 5% of the feature range is applied to characterize systematic biases during processing. Through this method, the training sample size is expanded to 300% of the original data. Finally, experiments show that this method significantly improves the model's generalization ability under unknown conditions while ensuring that the augmented data strictly adheres to physical constraint boundaries, fundamentally solving the overfitting risk caused by small sample training.

[0101] The MLP neural network constructed in this example adopts an LSTM-based core architecture. This network includes an input layer (for receiving multi-dimensional EMD feature vectors), three hidden layers, and corresponding memory cell vectors configured externally to effectively capture long-term dependencies in milling force signals. The output layer uses a linear activation function to directly generate signal sequence predictions. The network training uses the elastic backpropagation (Rprop) optimization algorithm, where the weights of each feature are initialized according to their correlation coefficients to enhance the physical interpretability of the model. The dynamic learning rate adjustment strategy is shown in Equation (8), which achieves adaptive parameter updates by segmenting the data and dividing it into time windows for iterative training, combined with the dynamic learning rate. This method not only effectively alleviates the gradient vanishing problem, but also improves the convergence speed by about 20% compared to traditional optimization algorithms. To verify the model's generalization ability, cross-validation is used in the validation phase to systematically evaluate the model's performance, thereby making a preliminary judgment on whether overfitting exists.

[0102] (8);

[0103] In the formula, : Current learning rate; T: Initial learning rate (0.6); P: Decreasing gradient (0.96); h: Current iteration number.

[0104] By dynamically monitoring key convergence metrics (validation loss rate, convergence speed, and spectral coherence) during the training iteration process, qualified models are added to the model pool in real time. During the testing phase, the optimal prediction model is selected based on a dual criterion—overfitting detection (training / validation loss ratio < 1.4) and test performance (MSE). Finally, ablation experiments and multi-model comparison experiments are conducted for the examples presented in this paper. The results of the key performance indicators are listed in Table 2.

[0105] Table 2 Performance Comparison of Milling Force Prediction Models and Ablation Experiment Results

[0106]

[0107] According to the experimental results in Table 2, the data preprocessing strategy and the selection of model architecture have a decisive impact on the accuracy of milling cutter force prediction, and the correct model architecture and effective preprocessing are the priority guarantees for multi-directional force prediction.

[0108] S60: Based on the oscillation boundary of the constraint force value of the radial cutting coefficient and the tangential cutting coefficient, the predicted milling force is adaptively filtered using a finite impulse response filter, and then the filtered predicted milling force is inversely normalized to finally output the multi-directional force prediction value.

[0109] The dynamic cutting force signal during milling contains rich information about the process state, but signal fitting methods face two major bottlenecks. First, the fitted signal suffers from redundant interference, containing a large amount of noise and non-critical harmonics, leading to model focusing distortion. Second, predicted values ​​often exceed the feasible region of material mechanics, losing their engineering practicality. To overcome these limitations, this invention proposes a signal post-processing framework based on feature engineering and physical constraints.

[0110] In this case, the main frequency range of the fitted signal is 20-80 Hz, with an overall oscillation range of [-10.75, +12.36], and the main frequency component oscillation range is [-6.42, +6.87]. After force signal fitting, inverse transform adjustment, and filtering, the final reconstructed signal is obtained: the mean square error after post-processing is 2.6913. The force signal fitting results are as follows. Figure 5 As shown.

[0111] In summary, this embodiment achieves the goal of directly predicting the multi-directional forces on the milling cutter under unknown working conditions from process parameters in a small sample size by using a process parameter-feature projection-signal fitting approach. In an exemplary embodiment, the overall flowchart of the final training model is as follows: Figure 6 As shown, it includes the following steps:

[0112] S1. Start: Initiate the model training process.

[0113] S2. Data preprocessing: The original milling force test signals are initially cleaned and organized.

[0114] S3, EMD decomposition: The signal is decomposed using the empirical mode decomposition method to extract the intrinsic mode functions (IMF).

[0115] S4. Data Augmentation: Expand training samples through strategies such as linear scaling and random translation to improve the model's generalization ability.

[0116] S5. Data Normalization: Standardize the enhanced data to eliminate the influence of dimensions.

[0117] S6. Feature Engineering: Extracting multi-dimensional features based on the decomposed signal.

[0118] S7. Feature Correlation Calculation: Calculate the Pearson correlation coefficient between each feature and the process parameters to quantify the strength of their correlation.

[0119] S8. Feature Dimensionality Reduction: Reduce feature dimensionality through methods such as Principal Component Analysis (PCA).

[0120] S9. Feature Filtering: Select features with a correlation coefficient higher than a set threshold as key features.

[0121] S10. Deep Formation of LSTM Network: Constructing a time-series prediction model based on Long Short-Term Memory (LSTM) network.

[0122] S11, Network Iteration n: The model is trained iteratively multiple times.

[0123] S12. Verify that the loss and test loss meet the requirements: Determine whether the current model has reached the preset performance indicators. If not (N), proceed to the next step; if yes (Y), merge the model into the model pool.

[0124] S13. Determine if the maximum number of iterations has been reached: Check if the maximum number of iterations has been reached. If yes (Y), end model training. If no (N), set n=n+1 and return to continue network iteration.

[0125] S14. Merge into the model pool: Store the trained model into the model pool.

[0126] S15. Select the best model: Select the best model from the model pool based on the validation loss and test performance (such as MSE).

[0127] S16. Post-processing model: Post-processing optimization of the prediction results under physical constraints.

[0128] S17. Recursive Optimization: Further optimize the model output through recursion.

[0129] S18. Filtering Model: An adaptive FIR filter is used to filter the predicted signal to suppress noise and non-critical harmonics.

[0130] S19. Final Output: Outputs the predicted value of the multi-directional milling force after post-processing and filtering.

[0131] This application also provides a device for predicting multi-directional forces during milling, such as... Figure 7 As shown, the multi-directional force prediction device during the milling process of the milling cutter includes:

[0132] The signal acquisition module 701 is configured to acquire a small sample of milling force signals based on the orthogonal test method, using rotational speed, feed rate, depth of cut and milling material as test factors;

[0133] The feature calculation module 702 is configured to perform time-frequency decomposition processing on the milling force signal using empirical mode decomposition to extract multi-dimensional features characterizing the dynamic characteristics of the milling force. The multi-dimensional features include the main frequency feature, peak force feature, total frequency energy feature, and main frequency amplitude.

[0134] The feature selection module 703 is configured to calculate the correlation between the engineering parameters corresponding to the experimental factors and the multi-dimensional features, wherein the correlation is quantified by the correlation coefficient, and features with a correlation coefficient greater than a set threshold are selected as key features.

[0135] The mapping model construction module 704 is configured to construct a physical mapping model from process parameters to the key features, and to solve for the radial cutting coefficient and the tangential cutting coefficient through the physical mapping model;

[0136] The milling prediction module 705 is configured to output the dynamic signal of the multi-directional force on the milling cutter as the predicted milling force based on the key features and through a time-varying signal prediction model. The time-varying signal prediction model is a time-series model based on a recurrent neural network.

[0137] The adaptive filtering module 706 is configured to adaptively filter the predicted milling force using a finite impulse response filter based on the oscillation boundary of the constraint force values ​​of the radial cutting coefficient and the tangential cutting coefficient, and then perform inverse normalization processing on the filtered predicted milling force to finally output the multi-directional force prediction value.

[0138] This application provides an electronic device. The electronic device may include a processor and a memory, wherein the processor and the memory can communicate; exemplarily, the processor and the memory communicate via a communication bus.

[0139] The processor executes computer execution instructions stored in memory, causing the processor to perform the scheme in the above embodiments. The processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0140] The communication bus can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. The system bus can be divided into address bus, data bus, control bus, etc. Transceivers are used to enable communication between database access devices and other computers (e.g., clients, read-write libraries, and read-only libraries). Memory may include random access memory (RAM) and may also include non-volatile memory.

[0141] The electronic device provided in this application embodiment can be the terminal device described in the above embodiments.

[0142] This application also provides a computer-readable storage medium storing computer instructions. When the computer instructions are executed on a computer, the computer performs the technical solution of the multi-directional force prediction method during milling as described above.

[0143] This application also provides a computer program product, which includes a computer program stored in a computer-readable storage medium. At least one processor can read the computer program from the computer-readable storage medium. When the at least one processor executes the computer program, it can implement the technical solution of the multi-directional force prediction method in the milling process described in the above embodiments.

[0144] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be indirect coupling or communication connection through some interfaces, devices, or modules, and may be electrical, mechanical, or other forms.

[0145] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to implement the solution of this embodiment according to actual needs.

[0146] Furthermore, the functional modules in the various embodiments of this application can be integrated into one processing unit, or each module can exist physically separately, or two or more modules can be integrated into one unit. The unit composed of the above modules can be implemented in hardware or in the form of hardware plus software functional units.

[0147] The integrated modules described above, implemented as software functional modules, can be stored in a computer-readable storage medium. These software functional modules, stored in a storage medium, include several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) or processor to execute some steps of the methods of the various embodiments of this application.

[0148] It should be understood that the aforementioned processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), etc. A general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in this invention can be directly manifested as execution by a hardware processor, or execution by a combination of hardware and software modules within the processor.

[0149] The memory may include high-speed RAM, and may also include non-volatile storage (NVM), such as at least one disk storage device, and may also be a USB flash drive, external hard drive, read-only memory, disk or optical disc, etc.

[0150] Buses can be Industry Standard Architecture (ISA) buses, Peripheral Component Interconnect (PCI) buses, or Extended Industry Standard Architecture (EISA) buses, etc. Buses can be categorized into address buses, data buses, control buses, etc.

[0151] The aforementioned storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. The storage medium can be any available medium that can be accessed by a general-purpose or special-purpose computer.

[0152] An exemplary storage medium is coupled to a processor, enabling the processor to read information from and write information to the storage medium. Alternatively, the storage medium can be an integral part of the processor. The processor and storage medium can reside in an Application Specific Integrated Circuit (ASIC). Alternatively, the processor and storage medium can exist as discrete components in an electronic control unit or main control device.

[0153] Those skilled in the art will understand that all or part of the steps of the above-described method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When executed, the program performs the steps of the above-described method embodiments; and the aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks.

[0154] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application.

Claims

1. A method for predicting multi-directional forces during milling, characterized in that, include: Based on the orthogonal test method, the milling force signal of a small sample is obtained by using the rotational speed, feed rate, depth of cut and milling material as test factors; The milling force signal is subjected to time-frequency decomposition processing using empirical mode decomposition to extract multi-dimensional features characterizing the dynamic properties of the milling force. These multi-dimensional features include dominant frequency features, peak force features, total frequency energy features, and dominant frequency amplitude. The correlation between the engineering parameters corresponding to the experimental factors and the multi-dimensional features is calculated. The correlation is quantified by the correlation coefficient, and features with a correlation coefficient greater than a set threshold are selected as key features. A physical mapping model is constructed to map process parameters to the key features, and the radial cutting coefficient and tangential cutting coefficient are solved through the physical mapping model. Based on the aforementioned key features, the dynamic signal of the multi-directional force on the milling cutter is output as the predicted milling force through a time-varying signal prediction model. The time-varying signal prediction model is a time-series model based on a recurrent neural network. Based on the oscillation boundary of the constraint force values ​​of the radial cutting coefficient and the tangential cutting coefficient, the predicted milling force is adaptively filtered using a finite impulse response filter, and then the filtered predicted milling force is inversely normalized to finally output the multi-directional force prediction value.

2. The method for predicting multi-directional forces during milling according to claim 1, characterized in that, The formula for calculating the correlation coefficient is: （1）； In the formula, r Represents the correlation coefficient. X i Indicates the engineering parameter value. Y i Represents eigenvalues; i Index representing the value of engineering parameter. k Indicates the total number of engineering parameters. This represents the average value of engineering parameters. This represents the average value of the eigenvalues.

3. The method for predicting multi-directional forces during milling according to claim 1, characterized in that, The milling force signal is subjected to time-frequency decomposition using empirical mode decomposition to extract multi-dimensional features characterizing the dynamic properties of the milling force, including: The main frequency characteristics are calculated using the following formula: （2）； In the formula, f dom Main frequency characteristics; c For frequency coefficients; Z Number of teeth; This refers to the vibration offset related to the rotational speed. n The current rotational speed; a p For depth of cut; Peak force characteristics are calculated using the following formula: （3）； In the formula, F peak Characterized by peak force. K c The material-related cutting coefficient; α The material shear strain rate; β For tool-chip friction; f For feed rate; The total frequency energy characteristic is calculated using the following formula: （4）； In the formula, E total The total frequency energy characteristic; C is the energy conversion parameter; C is the total frequency energy coefficient. The main frequency amplitude is calculated using the following formula: （5）； In the formula, Amplitude minus cutting coefficient; Main frequency amplitude; Z Number of teeth; It is a non-linear exponent.

4. The method for predicting characteristic multi-directional forces during milling as described in claim 1, characterized in that, The physical mapping model includes the radial and tangential micro-element cutting force models of the milling cutter, which are expressed as follows: （6）； （7）； In the formula, F t For tangential cutting force, F r Radial cutting force, The tangential cutting coefficient is... h ( θ () represents the instantaneous, undeformed chip thickness. dz The height of the axial micro-element, The first cutting edge force coefficient, The radial cutting coefficient is... This is the second cutting edge force coefficient.

5. The method for predicting multi-directional forces during milling according to claim 1, characterized in that, The time-varying signal prediction model employs a dynamic learning rate adjustment strategy during training, and the learning rate calculation formula satisfies: （8）； In the formula, is the current learning rate; T is the initial learning rate; P is the decaying gradient; h is the current iteration number.

6. The method for predicting multi-directional forces during milling according to claim 1, characterized in that, The time-varying signal prediction model is an LSTM network, which includes an input layer, three hidden layers, and an output layer. The input layer receives a multi-dimensional feature vector obtained through empirical mode decomposition. The hidden layer is configured with memory cell vectors to capture the long-term dependence of the milling force signal. The output layer uses a linear activation function to generate a dynamic signal sequence of multi-directional forces on the milling cutter. The LSTM network is trained using an elastic backpropagation optimization algorithm, and the initial weights of each feature are set according to their correlation coefficient with engineering parameters.

7. The method for predicting multi-directional forces during milling according to claim 1, characterized in that, The oscillation range of the finite impulse response filter is dynamically adjusted based on the obtained radial and tangential cutting coefficients.

8. A device for predicting multi-directional forces during milling, characterized in that, The device includes: The signal acquisition module is configured to acquire a small sample of milling force signals based on the orthogonal experimental method, using rotational speed, feed rate, depth of cut, and milling material as experimental factors; The feature calculation module is configured to perform time-frequency decomposition processing on the milling force signal using empirical mode decomposition, and extract multi-dimensional features characterizing the dynamic characteristics of the milling force. The multi-dimensional features include the main frequency feature, peak force feature, total frequency energy feature, and main frequency amplitude. The feature filtering module is configured to calculate the correlation between the engineering parameters corresponding to the experimental factors and the multi-dimensional features. The correlation is quantified by the correlation coefficient, and features with a correlation coefficient greater than a set threshold are selected as key features. The mapping model construction module is configured to construct a physical mapping model from process parameters to the key features, and solve for the radial cutting coefficient and tangential cutting coefficient through the physical mapping model; The milling prediction module is configured to output the dynamic signal of the multi-directional force on the milling cutter as the predicted milling force based on the key features and through a time-varying signal prediction model. The time-varying signal prediction model is a time-series model based on a recurrent neural network. The adaptive filtering module is configured to adaptively filter the predicted milling force using a finite impulse response filter based on the oscillation boundary of the constraint force values ​​of the radial cutting coefficient and the tangential cutting coefficient, and then perform inverse normalization processing on the filtered predicted milling force to finally output the multi-directional force prediction value.

9. An electronic device, characterized in that, include: A processor, and a memory communicatively connected to the processor; The memory stores computer-executed instructions; The processor executes the computer execution instructions stored in the memory to implement the method for predicting multi-directional forces during milling as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the method for predicting multi-directional forces during milling as described in any one of claims 1-7.

Images