Milling stability prediction method and system of machine tool based on hybrid drive

Abstract

The invention discloses a machine tool milling stability prediction method and system based on hybrid drive, and belongs to the technical field of advanced manufacturing device prediction. Then, reverse training is conducted on the prediction network through a reverse propagation algorithm by means of distribution difference loss to complete domain adaptation of a simulation domain and a test domain, and therefore a milling chatter stability prediction model is obtained, and the milling stability of the milling machine is predicted according to the milling machine machining working condition parameters by means of the obtained milling chatter stability prediction model. The problems of nonlinearity and uncertainty can be more effectively solved, and flutter stability domain deviation caused by nonlinearity of a spindle system and uncertainty of model parameters can be effectively handled. The local maximum mean value difference index can extract the distribution difference in the simulation flutter data and the test flutter data, and the association relationship information contained in the data is fully extracted.

Description

Technical Field

[0001] The present invention belongs to the technical field of advanced manufacturing device prediction, and in particular relates to a method and system for predicting machine tool milling stability based on hybrid drive. Background Art

[0002] With the rapid development of modern industry, the surface quality requirements of parts in the manufacturing fields of automobiles, ships, aerospace, aviation, etc. are becoming more and more stringent, the complexity of part structures is becoming higher and higher, and the requirements for the processing capabilities of CNC machine tools are also constantly increasing. The selection of cutting parameters in the processing process must match the corresponding requirements of the parts themselves on the one hand, and on the other hand, it must also be adapted to the cutting process. For example, in actual processing, the selection of processing parameters is often too conservative, making it difficult to give full play to the working efficiency of the machine tool and limiting the reduction of costs; in addition, if high-efficiency processing parameters are blindly selected, it is easy to cause chatter in the cutting process, which seriously affects the processing quality of parts, aggravates tool wear, and increases the risk of equipment failure. Self-excited chatter is still the main factor limiting milling productivity in industrial environments, and the growing demand has magnified the impact of this factor, becoming a bottleneck problem restricting my country's manufacturing industry from moving towards high-precision and high-tech.

[0003] In the field of advanced manufacturing, it is very important to avoid the occurrence of chatter. In order to achieve this goal, chatter stability domain prediction technology has emerged. The traditional technical means of predicting chatter stability domain is mechanism modeling. By constructing a machine tool dynamics model, a cutting mechanics model, and a milling dynamics model, the milling stability domain composed of rotation speed and critical cutting depth can be predicted, which can effectively optimize the selection process of machining parameters, ensure machining accuracy while improving machining efficiency, and thus achieve high-performance machining of CNC machining equipment. However, with the continuous improvement of manufacturing needs, this requirement is becoming increasingly difficult to meet actual engineering needs. The main problem is that there is a large difference between the chatter stability domain predicted by the mechanism model and the stability domain in actual engineering processing, which misleads the selection of cutting parameters and cannot effectively avoid chatter. The reasons for this difference are mainly from two aspects: (1) The factors inducing the chatter process are extremely complex. The chatter dynamics model itself contains many assumptions and it is difficult to cover all the chatter mechanisms and influencing factors. This leads to an inevitable essential difference between the chatter dynamics model and the actual chatter physical object process. This difference is particularly obvious in high-precision manufacturing processes. (2) The chatter dynamics model involves many input parameters, including the structural geometric parameters of the tool and the tool holder, the material parameters of the tool and the workpiece, the dynamic characteristic parameters including the natural frequency and damping ratio, and the cutting force coefficient, etc. On the one hand, these parameters are uncertain and difficult to accurately identify. On the other hand, the machine tool dynamic characteristic parameters and the cutting force coefficient will change with the changes in the processing conditions. This change is more significant in high-demand processing scenarios. These two reasons limit the application of the chatter mechanism model in actual engineering.

[0004] In recent years, with the rapid development of artificial intelligence technology, especially thanks to the significant advantages of machine learning technology in processing nonlinear features and identifying system changes, data-driven intelligent chatter stability analysis has received more and more attention. However, such methods usually require a large number of chatter samples in experiments or tests. In order to solve this problem, mechanism data fusion or hybrid drive methods have been proposed and have been increasingly applied. There are two main types of such hybrid drive methods. One is the probabilistic method represented by Bayesian reasoning. This method classifies all uncertainties and changes as probabilistic uncertainty problems and obtains a probabilistic chatter stability domain. However, this view cannot effectively deal with the nonlinear factors in the chatter stability domain; the other method is the transfer learning method, which extracts information from the mechanism or simulation model to reduce the requirements for the amount of experimental or test data. By using actual test data in the modeling process, this has greatly improved the defects of traditional milling stability prediction methods.

[0005] The main bottleneck of the current intelligent stability domain prediction method based on transfer learning is the efficiency problem caused by insufficient data. Specifically, a large number of experiments are required, or the mechanism model must be repeatedly called in each training to generate sufficient data, or additional manual operations or feature engineering are required to deal with data sparsity problems, such as introducing sufficient randomness by giving uncertain ranges to parameters and building sub-networks to achieve ensemble learning. These operations require a lot of manual intervention, which greatly limits the application of the method in practical engineering. Therefore, there is an urgent need for an intelligent milling stability prediction model that can be adaptively trained to improve application efficiency. Summary of the invention

[0006] The purpose of the present invention is to provide a method and system for predicting machine tool milling stability based on hybrid drive, so as to overcome the problem of poor milling stability prediction accuracy in the prior art. The method of the present invention improves the prediction performance while maintaining a high implementation efficiency.

[0007] To achieve the above purpose, the technical solution adopted by the present invention is as follows: A method for predicting milling stability of a machine tool based on hybrid drive comprises the following steps: S1, generating simulation data set based on milling stability prediction mechanism model and machining condition parameters; S2, performing a chatter stability test on the milling process of the machine tool and obtaining the milling process signal, and constructing a test data set based on the milling process signal; S3, using the deep subdomain adaptation network model to extract features from the simulation dataset and the test dataset for classification and prediction of milling chatter stability state; S4, calculating the distribution difference between the simulation data and the flutter test data according to the extracted features in the simulation data set and the test data set, and calculating the distribution difference loss based on the distribution difference; S5, the prediction network is trained using the simulation data set and the test data set, and then the prediction network is reversely trained through the back propagation algorithm using the distribution difference loss to complete the domain adaptation of the simulation domain and the test domain, thereby obtaining the milling chatter stability prediction model, and the milling stability of the milling machine is predicted according to the milling machine processing condition parameters using the obtained milling chatter stability prediction model.

[0008] Preferably, the milling stability prediction mechanism model includes a milling cutting mechanics model, a machine tool processing dynamics model and a chatter prediction model.

[0009] Preferably, the simulation data set constitutes the source domain, which is expressed as , which contains labeled simulation samples; input For cutting condition parameters, input parameter definition ,in , , and They represent the spindle speed, axial cutting depth, radial cutting depth and the clamping length of the tool in the tool holder. All optional working conditions together constitute the feature space .

[0010] Preferably, a data acquisition system and a sensor system are used to collect signals during the milling process of a machine tool, and the collected signals are processed by signal processing technology to obtain a signal spectrum, so as to determine whether the stability state, i.e., chatter occurs, and a test data set is constructed based on the determination result.

[0011] Preferably, a local maximum mean distribution difference LMMD evaluation index is constructed to evaluate the distribution difference between the simulation data and the flutter test data.

[0012] Preferably, the distribution difference loss calculated based on the local maximum mean difference of the simulation and test samples of the vibration, and the classification loss of the model together constitute the total loss of the model training.

[0013] Preferably, the distribution difference loss is expressed as:

[0014] Where L is the set of all domain adaptation layers; Classification loss of Deep Sub-Domain Adaptation Network DSAN;

[0015] in, is the categorical cross entropy loss, is the total number of simulation and test samples, and are sample features and corresponding labels.

[0016] A machine tool milling stability prediction system based on hybrid drive, comprising a simulation module, a test module, a classification prediction module, a difference loss module and a prediction module; Simulation module, which generates simulation data sets based on the milling stability prediction mechanism model and machining condition parameters; The test module performs a chatter stability test on the milling process of the machine tool and obtains the milling process signal, and constructs a test data set based on the milling process signal; The classification prediction module uses a deep subdomain adaptation network model to extract features from the simulation data set and the test data set for classification prediction of milling chatter stability state; A difference loss module calculates the distribution difference between the simulation data and the flutter test data according to the extracted features in the simulation data set and the test data set, and calculates the distribution difference loss based on the distribution difference; The prediction module uses the simulation data set and the test data set to train the prediction network, and then uses the distribution difference loss to reversely train the prediction network through the back propagation algorithm to complete the domain adaptation of the simulation domain and the test domain, so as to obtain the milling vibration stability prediction model. The obtained milling vibration stability prediction model is used to predict the milling stability of the milling machine according to the milling machine processing condition parameters.

[0017] A computer device comprises a memory, a processor and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the above-mentioned hybrid drive-based intelligent milling stability prediction method are implemented.

[0018] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the above-mentioned hybrid drive intelligent milling stability prediction method. Compared with the prior art, the present invention has the following beneficial technical effects: The present invention provides a method for predicting the milling stability of a machine tool based on hybrid drive, which uses a simulation data set and a test data set to train a prediction network, and then uses a distribution difference loss to reversely train the prediction network through a back propagation algorithm to complete the domain adaptation of the simulation domain and the test domain, thereby obtaining a milling vibration stability prediction model, and using the obtained milling vibration stability prediction model to predict the milling stability of the milling machine according to the milling machine processing condition parameters, which can more effectively handle nonlinear and uncertainty problems, and can effectively deal with the vibration stability domain deviation caused by the nonlinearity of the spindle system and the uncertainty of the model parameters. The local maximum mean difference index can extract the distribution difference in the simulation vibration data and the test vibration data, and fully extract the correlation information contained in the data.

[0019] Preferably, the loss function of the feedforward neural network is expanded based on the local maximum mean difference. Combined with the back propagation algorithm, the algorithm automatically completes the distribution difference evaluation and shrinkage during the iteration process, realizes end-to-end training, avoids additional operations such as human intervention and feature engineering, has high analysis efficiency, and has good practical application prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 It is a flow chart of the hybrid drive-based intelligent milling stability prediction method in an embodiment of the present invention.

[0021] Figure 2 It is the deep sub-domain adaptation network model in the embodiment of the present invention.

[0022] Figure 3 It is a schematic diagram of the subdomain adaptation process in an embodiment of the present invention.

[0023] Figure 4 1 is a diagram of the prediction result and experimental result of the flutter stability in an embodiment of the present invention.

[0024] Figure 5 It is the experimental result based on the sample quantity dependence experiment in the embodiment of the present invention. DETAILED DESCRIPTION

[0025] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0026] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0027] like Figure 1 As shown, the present invention provides a method for predicting the milling stability of a machine tool based on hybrid drive, which is used to effectively detect the stability of hybrid drive intelligent milling, and specifically includes the following steps: S1, generating simulation data set based on milling stability prediction mechanism model and machining condition parameters; S2, performing a chatter stability test on the milling process of the machine tool and obtaining the milling process signal, and constructing a test data set based on the milling process signal; S3, using the deep subdomain adaptation network model to extract features from the simulation dataset and the test dataset for classification and prediction of milling chatter stability state; S4, calculating the distribution difference between the simulation data and the flutter test data according to the extracted features in the simulation data set and the test data set, and calculating the distribution difference loss based on the distribution difference; S5, the prediction network is trained using the simulation data set and the test data set, and then the prediction network is reversely trained through the back propagation algorithm using the distribution difference loss to complete the domain adaptation of the simulation domain and the test domain, thereby obtaining the milling chatter stability prediction model, and the milling stability of the milling machine is predicted according to the milling machine processing condition parameters using the obtained milling chatter stability prediction model.

[0028] In the specific implementation of the present application, the milling stability prediction mechanism model includes a milling cutting mechanics model, a machine tool processing dynamics model and a chatter prediction model. According to the definition of transfer learning, the simulation data set constitutes the source domain, which is expressed as , which contains Labeled simulation samples. Input For cutting condition parameters, input parameter definition ,in , , and They represent the spindle speed, axial cutting depth, radial cutting depth and the clamping length of the tool in the tool holder. All optional working conditions together constitute the feature space In this paper, milling stability analysis is considered as a classification problem, so the label space is defined as , where 0 indicates that the milling process is in a stable state, and 1 indicates that the milling process is in a chattering state. Yes Input The corresponding label.

[0029] Specifically, the chatter stability test is performed on the milling process of the machine tool: the data acquisition system and the sensor system are used to collect signals during the milling process of the machine tool. The collected signals can be acoustic signals or cutting force signals. The collected signals are processed by signal processing technology to obtain signal spectra, based on which the stability state, i.e., whether chatter occurs, can be determined, and a test data set is constructed based on the judgment results. The signal processing technology specifically uses the fast Fourier transform method to transform the collected signals into signal spectra.

[0030] According to the definition of transfer learning, the test dataset is defined as the target domain , which contains labeled test samples. Yes Input The corresponding label. The feature space of the target domain is , contains the same features as the feature space of the source domain, and assumes that the target domain and the source domain share the same label space.

[0031] The deep subdomain adaptation network model is used to extract features from simulation data and test data: the model structure of the deep subdomain adaptation network DSAN is referenced Figure 2; Based on the general deep network model structure, this network model adds the last few layers of domain adaptation layers to align the source domains extracted from these hidden layers and target domain Features can shrink the data distribution of the source domain and the target domain and extract information in the source domain and the target domain more effectively.

[0032] The distribution difference between the simulation data and the flutter test data is calculated according to the extracted features in the simulation data set and the test data set. Specifically, a local maximum mean distribution difference (LMMD) evaluation index is constructed to evaluate the distribution difference between the simulation data and the flutter test data.

[0033] 4.1) The calculation process of the local maximum mean distribution difference LMMD between simulation data and flutter test data is as follows: The calculation formula of the local maximum mean difference is as follows:

[0034] Where p and q represent the source domains respectively. and target domain The marginal probability distribution of and yes and In the present invention, the subdomain is defined as It is divided according to the label category. Since milling stability has two categories: stable 0 and chatter 1, and Each contains two subdomains. and Is a subdomain and The marginal probability distribution of . It means that the sample features are mapped to the regenerated Hilbert RKHS space, which has a feature kernel function k. Represents some feature maps that can map the original samples into the RKHS space. represents the mathematical expectation of category c.

[0035] The local maximum mean difference can evaluate the local distribution difference between simulation and chatter test data in stable class samples and chatter class samples respectively. By minimizing this indicator in the deep network, the distribution difference of related subdomains within the same category is reduced.

[0036] The characteristic kernel function k of the regenerated Hilbert RKHS space can be expressed as ,in Represents the vector inner product.

[0037] The local maximum mean difference can be estimated as follows

[0038] in and Represents the simulation samples and test samples The calculation process of sample weight is as follows:

[0039] therefore and For simulation or test samples, the expression They all represent the weighted sum of samples for category c.

[0040] In the domain adaptation layer l, the features of the simulation samples and test samples extracted by the neurons are and Using the kernel function k, the local maximum mean difference between the simulation and test samples can be directly calculated by the following formula: Substituting the features extracted from the hidden layer into the calculation formula, the local maximum mean difference LMMD of the simulation and test samples can be obtained.

[0041] The total difference between simulation data and test samples is based on LMMD, and multi-kernel local maximum mean difference MK-LMMD is used to help more fully extract distribution difference features. MK-LMMD is implemented by taking multiple Gaussian kernel functions with different parameters.

[0042] 4.2) The distribution difference loss is calculated based on the local maximum mean difference of the simulation and test samples of the vibration, and the classification loss of the model together constitutes the total loss of the model training.

[0043] The distribution difference loss is expressed as

[0044] where L is the set of all domain adaptation layers.

[0045] Classification loss for Deep Subdomain Adaptation Network (DSAN).

[0046]

[0047] in, is the categorical cross entropy loss, is the total number of simulation and test samples, and are sample features and corresponding labels.

[0048] The prediction model is trained using the simulation data set and the test data set; first the samples are normalized, and then the normalized samples are used for training: 5.1) Through pre-training, the network model learns the basic associations in the flutter prediction and analysis task. Input simulation sample set , train and obtain the weight matrix W. Fix the weights W1~WN1 of the first N1 layers of the network, corresponding to the common features of the prediction task.

[0049] 5.2) Further, using the flutter test sample set Fine-tune the model. Based on the weight matrix W, input the test sample set , get the updated weight matrix, and then fix the network weights WN1+1~WN1+N2 of the N1+1~N1+N2 layers. This part of the network still extracts relatively preliminary features.

[0050] 5.3) N3 training is completed using the distribution difference loss and back propagation algorithm. The optimization objective function of DSAN used to train the chatter prediction analysis is:

[0051] The training process uses the adaptive motion estimation algorithm Adam. The principle of the subdomain adaptation process is referenced Figure 3 This process does not require manual operation and can be automatically calculated during the iteration process.

[0052] The test samples are input into the model to verify the validity of the model. The present invention is further described below in conjunction with a specific flutter prediction case.

[0053] A vertical machining center is used, the spindle speed range is [5000 rpm, 11000 rpm], the radial cutting depth range is [50%*D, D], where D is the diameter of the tool, and the clamping length range of the tool is [25 mm, 45 mm]. The vibration stability domain diagram depends on the above three parameters, and a total of 15 combinations are randomly selected to form 15 cutting cases. In each case, 25 different speed values ​​(interval of 250 rpm) are uniformly selected, and 5 uniform axial cutting depths are selected, so a single case has 125 sample points. A total of 1875 test samples are tested.

[0054] For this sample set, 400 sample points are randomly selected to form training sets of different sizes, and 1441 sample points are selected from these 100 sample points as test sets.

[0055] First, 100 sample points are randomly selected to form a training set for training. Figure 4The test and prediction results with a clamping length of 25 mm and a radial cutting depth of 0.55*D are given. The prediction results include the flutter stability boundary of the mechanism model and the boundary predicted by the method proposed in the present invention. It can be seen that the method proposed in the present invention can significantly improve the accuracy of the prediction results. The prediction accuracy is defined as the ratio of the number of samples with correct stability prediction to the total number of samples in the verification set. According to the verification results of the complete test set, the prediction accuracy of the mechanism model is 64.1%, while the prediction accuracy of the method of the present invention is 84.04%.

[0056] Figure 5 This is a schematic diagram of the verification accuracy achieved by this method using training sample sets of different sizes. It can be seen that when the number of training samples increases, this method still has better performance than other methods, and the prediction accuracy will increase with the increase of training samples.

[0057] Through the above embodiments and experimental results, it can be found that the invention can effectively predict the chatter stability state of the processing condition, and help to effectively avoid the occurrence of chatter during the processing.

[0058] A machine tool milling stability prediction system based on hybrid drive, comprising a simulation module, a test module, a classification prediction module, a difference loss module and a prediction module; Simulation module, which generates simulation data sets based on the milling stability prediction mechanism model and machining condition parameters; The test module performs a chatter stability test on the milling process of the machine tool and obtains the milling process signal, and constructs a test data set based on the milling process signal; The classification prediction module uses a deep subdomain adaptation network model to extract features from the simulation data set and the test data set for classification prediction of milling chatter stability state; A difference loss module calculates the distribution difference between the simulation data and the flutter test data according to the extracted features in the simulation data set and the test data set, and calculates the distribution difference loss based on the distribution difference; The prediction module uses the simulation data set and the test data set to train the prediction network, and then uses the distribution difference loss to reversely train the prediction network through the back propagation algorithm to complete the domain adaptation of the simulation domain and the test domain, so as to obtain the milling vibration stability prediction model. The obtained milling vibration stability prediction model is used to predict the milling stability of the milling machine according to the milling machine processing condition parameters.

[0059] In another embodiment of the present invention, a terminal device is provided, the terminal device includes a processor and a memory, the memory is used to store a computer program, the computer program includes program instructions, and the processor is used to execute the program instructions stored in the computer storage medium. The processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc., which are the computing core and control core of the terminal, which are suitable for implementing one or more instructions, and are specifically suitable for loading and executing one or more instructions to implement the corresponding method flow or corresponding functions; the processor described in the embodiment of the present invention is based on the operation of the hybrid drive intelligent milling stability prediction method.

[0060] In another embodiment of the present invention, the present invention also provides a storage medium, specifically a computer-readable storage medium (Memory), which is a memory device in a terminal device for storing programs and data. It is understandable that the computer-readable storage medium here can include both the built-in storage medium in the terminal device and the extended storage medium supported by the terminal device. The computer-readable storage medium provides a storage space, which stores the operating system of the terminal. In addition, one or more instructions suitable for being loaded and executed by the processor are also stored in the storage space, and these instructions can be one or more computer programs (including program codes). It should be noted that the computer-readable storage medium here can be a high-speed RAM memory or a non-volatile memory, such as at least one disk memory. The processor can load and execute one or more instructions stored in the computer-readable storage medium to implement the corresponding steps of the above-mentioned embodiment of the method for predicting the stability of intelligent milling based on hybrid drive.

[0061] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0062] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0063] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0064] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0065] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for predicting machine tool milling stability based on hybrid drive, characterized in that: The following steps are involved: S1, generating simulation data set based on milling stability prediction mechanism model and machining condition parameters; S2, performing a chatter stability test on the milling process of the machine tool and obtaining the milling process signal, and constructing a test data set based on the milling process signal; S3, using the deep subdomain adaptation network model to extract features from the simulation dataset and the test dataset for classification and prediction of milling chatter stability state; S4, calculating the distribution difference between the simulation data and the flutter test data according to the extracted features in the simulation data set and the test data set, and calculating the distribution difference loss based on the distribution difference; S5, the prediction network is trained using the simulation data set and the test data set, and then the prediction network is reversely trained through the back propagation algorithm using the distribution difference loss to complete the domain adaptation of the simulation domain and the test domain, thereby obtaining the milling chatter stability prediction model, and the milling stability of the milling machine is predicted according to the milling machine processing condition parameters using the obtained milling chatter stability prediction model.

2. The method for predicting milling stability of a machine tool based on hybrid drive according to claim 1, characterized in that: The milling stability prediction mechanism model includes a milling cutting mechanics model, a machine tool processing dynamics model and a chatter prediction model.

3. The method for predicting milling stability of a machine tool based on hybrid drive according to claim 2, characterized in that: The simulated data set constitutes the source domain, expressed as , which contains labeled simulation samples; input For cutting condition parameters, input parameter definition ,in , , and They represent the spindle speed, axial cutting depth, radial cutting depth and the clamping length of the tool in the tool holder. All optional working conditions together constitute the feature space .

4. The method for predicting milling stability of a machine tool based on hybrid drive according to claim 1, characterized in that: The data acquisition system and sensor system are used to collect signals during the milling process of the machine tool. The collected signals are processed through signal processing technology to obtain the signal spectrum. The stability state, that is, whether chatter occurs, can be determined, and a test data set is constructed based on the judgment results.

5. The method for predicting milling stability of a machine tool based on hybrid drive according to claim 1, characterized in that: The local maximum mean distribution difference (LMMD) evaluation index is constructed to evaluate the distribution difference between simulation data and flutter test data.

6. The method for predicting milling stability of a machine tool based on hybrid drive according to claim 5, characterized in that: The distribution difference loss calculated based on the local maximum mean difference of the simulation and test samples of Flutter, and the classification loss of the model together constitute the total loss of model training.

7. The method for predicting milling stability of a machine tool based on hybrid drive according to claim 6, characterized in that: The distribution difference loss is expressed as: Where L is the set of all domain adaptation layers; Classification loss of Deep Sub-Domain Adaptation Network DSAN; in, is the categorical cross entropy loss, is the total number of simulation and test samples, and are sample features and corresponding labels.

8. A machine tool milling stability prediction system based on hybrid drive, characterized in that: It includes simulation module, test module, classification prediction module, difference loss module and prediction module; Simulation module, which generates simulation data sets based on the milling stability prediction mechanism model and machining condition parameters; The test module performs a chatter stability test on the milling process of the machine tool and obtains the milling process signal, and constructs a test data set based on the milling process signal; The classification prediction module uses a deep subdomain adaptation network model to extract features from the simulation data set and the test data set for classification prediction of milling chatter stability state; A difference loss module calculates the distribution difference between the simulation data and the flutter test data according to the extracted features in the simulation data set and the test data set, and calculates the distribution difference loss based on the distribution difference; The prediction module uses the simulation data set and the test data set to train the prediction network, and then uses the distribution difference loss to reversely train the prediction network through the back propagation algorithm to complete the domain adaptation of the simulation domain and the test domain, so as to obtain the milling vibration stability prediction model. The obtained milling vibration stability prediction model is used to predict the milling stability of the milling machine according to the milling machine processing condition parameters.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the hybrid drive-based intelligent milling stability prediction method as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the hybrid drive-based intelligent milling stability prediction method as claimed in any one of claims 1 to 7 are implemented.

Images