A method and system for predicting instantaneous cutting force with physical information embedding

Abstract

The application belongs to the technical field of cutting force prediction, and discloses a physical information embedded instantaneous cutting force prediction method and system. The method comprises the following steps: inputting machining time series data into a feature extraction network of a trained instantaneous cutting force prediction model to obtain features of the machining time series data; inputting static machining parameters into a theoretical mechanical cutting force model to calculate corresponding theoretical cutting forces; and obtaining a mechanism feature vector by splicing the theoretical cutting forces and the static machining parameters, inputting the mechanism feature vector into a full connection layer of the instantaneous cutting force prediction model; in the full connection layer, the mechanism feature vector is spliced with the obtained machining time series data features and is regressed to obtain an instantaneous cutting force prediction result. The application can improve the accuracy of cutting force prediction, improve the model generalization ability, improve the real-time performance of prediction, and increase the model interpretability.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of cutting force prediction, and more particularly relates to a physical information embedded instantaneous cutting force prediction method and system. BACKGROUND

[0002] Boring is a common method for manufacturing deep grooves or deep cuts, which involves the axial movement of the tool into the workpiece and the use of rotating tools to remove material. This technology is known for axial feed, which reduces cutting force, improves chip removal, and extends tool life. In boring machining, accurate prediction of cutting force is of great significance for optimizing machining parameters, designing robust machining processes, and preventing tool failure.

[0003] In the complex boring process, direct measurement and collection of cutting force is very difficult, and the cost of intelligent tool holder that can detect cutting force is very high, therefore, the cutting force of boring machining is generally predicted by related data in the machining process. With the rise of deep learning, cutting force prediction based on deep neural networks occupies an increasingly important position.

[0004] Existing cutting force prediction based on deep learning mainly adopts single data detection method, such as cutting force prediction based on current signal or vibration signal in the machining process, which leads to insufficient prediction accuracy and limited generalization. At the same time, the existing prediction model also has the problem of poor interpretability. Moreover, many prediction models are limited by large computational load and usually use cloud computing for real-time diagnosis, thereby increasing the dependence on the network, and when the network is disconnected, it cannot work normally, which seriously affects the efficiency of production. SUMMARY

[0005] In view of the above defects or improvement needs of the prior art, the present application provides a physical information embedded instantaneous cutting force prediction method and system, which aims to improve the accuracy of cutting force prediction, improve the model generalization ability, and improve the real-time performance of prediction.

[0006] To achieve the above-mentioned purpose, according to the first aspect of the present application, a physical information embedded instantaneous cutting force prediction method is provided, comprising:

[0007] inputting the machining time series data into the feature extraction network of the trained instantaneous cutting force prediction model to obtain the features of the machining time series data; wherein the machining time series data includes tool vibration signal and machine tool motor current signal;

[0008] The static machining parameters are input into a theoretical mechanical cutting force model to calculate corresponding theoretical cutting forces; and a mechanism feature vector is obtained by splicing the theoretical cutting forces and the static machining parameters, and the mechanism feature vector is input into a full connection layer of the instantaneous cutting force prediction model; wherein the static machining parameters include a rotational speed of a tool, a feed rate and a cutting depth;

[0009] In the full connection layer, the mechanism feature vector is spliced with acquired machining time sequence data features and is regressed to obtain an instantaneous cutting force prediction result.

[0010] Further, the features of the machining time sequence data include global time sequence features and local time sequence features;

[0011] The feature extraction network includes:

[0012] an embedding layer configured to convert the machining time sequence data into a high-dimensional time sequence vector;

[0013] a position encoding layer configured to perform position encoding on the high-dimensional time sequence vector to obtain encoded machining time sequence data;

[0014] an attention layer configured to calculate weights of features of different scales of the encoded machining time sequence data by using an attention mechanism, and to perform weighted summation on the features of different scales by using the corresponding weights to obtain fused global time sequence features;

[0015] a feedforward layer configured to perform nonlinear mapping on the global time sequence features;

[0016] a normalization layer configured to normalize the global time sequence features after the nonlinear mapping;

[0017] a convolution module configured to perform local feature extraction on the normalized global time sequence features to obtain more fine-grained local time sequence features;

[0018] Correspondingly, the full connection layer is configured to splice and regress the normalized global time sequence features, the local time sequence features and the mechanism feature vector to obtain an instantaneous cutting force prediction result.

[0019] Further, the theoretical mechanical cutting force model is:

[0020]

[0021] wherein, F c is a theoretical cutting force, A is a yield strength, B is a strain hardening coefficient, C is a strain rate sensitivity coefficient, n is a strain hardening index, m is a thermal softening index, ∈ is a shear plastic strain; v is a relative speed between a tool and a workpiece material, and is determined based on a rotational speed of the tool; h is an undeformed chip thickness, T is a temperature of the workpiece material, and Troom T is room temperature melt T is melting point of workpiece material; φ c β is shear angle n α is friction angle n a is rake angle of tool p f is cutting depth; f is feed rate.

[0022] Further, the theoretical mechanical cutting force model is constructed in the following way:

[0023] The theoretical cutting force F is established based on Merchant's cutting theory c The relationship between the theoretical cutting force F and the shear force on the shear surface is:

[0024]

[0025] In the formula, τ s A is shear yield strength of the machined material s A is shear area

[0026] The shear yield strength τ is replaced by the flow stress σ determined based on the simplified Johnson-Cook constitutive model s ; wherein the flow stress σ is:

[0027]

[0028] The shear area A s is determined by the cutting depth a p and the feed rate f:

[0029] A s = a p f sin φ c

[0030] The calculation formula of the shear yield strength τ s and the shear area A s is brought into the relationship between the theoretical cutting force F c and the shear force on the shear surface, to obtain the theoretical mechanical cutting force model.

[0031] Further, the attention mechanism is a multi-scale attention mechanism.

[0032] According to the second aspect of the present application, a physical information embedded instantaneous cutting force prediction system is provided, comprising: a trained instantaneous cutting force prediction model and a mechanism feature vector acquisition module; the instantaneous cutting force prediction model comprises a feature extraction network and a full connection layer;

[0033] The feature extraction network is used to extract features of the machining time series data; wherein the machining time series data comprises a tool vibration signal and a machine tool motor current signal;

[0034] The mechanism feature vector acquisition module is configured to input static machining parameters into a theoretical mechanical cutting force model, calculate a corresponding theoretical cutting force, and splice the theoretical cutting force with the static machining parameters to obtain a mechanism feature vector; wherein the static machining parameters comprise a tool rotation speed, a feed rate, and a cutting depth.

[0035] The fully connected layer is configured to splice the mechanism feature vector with the acquired machining time series data features and regress to obtain an instantaneous cutting force prediction result.

[0036] Further, the features of the machining time series data comprise global time series features and local time series features.

[0037] The feature extraction network comprises:

[0038] An embedding layer is configured to convert the machining time series data into a high-dimensional time series vector.

[0039] A position encoding layer is configured to perform position encoding on the high-dimensional time series vector to obtain encoded machining time series data.

[0040] An attention layer is configured to calculate weights of features of different scales of the encoded machining time series data by using an attention mechanism, and perform weighted summation on features of different scales by using corresponding weights to obtain fused global time series features.

[0041] A feedforward layer is configured to perform nonlinear mapping on the global time series features.

[0042] A normalization layer is configured to normalize the global time series features after nonlinear mapping.

[0043] A convolution module is configured to perform local feature extraction on the normalized global time series features to obtain more fine-grained local time series features.

[0044] Correspondingly, the fully connected layer is configured to splice the normalized global time series features, the local time series features, and the mechanism feature vector and regress to obtain an instantaneous cutting force prediction result.

[0045] According to a third aspect of the present application, an electronic device is provided, comprising a computer readable storage medium and a processor.

[0046] The computer readable storage medium is configured to store executable instructions.

[0047] The processor is configured to read executable instructions stored in the computer readable storage medium to execute the physical information embedded instantaneous cutting force prediction method of any one of the first aspect.

[0048] According to a fourth aspect of the present application, a computer readable storage medium is provided, having stored thereon a computer program which, when executed by a processor, implements the physical information embedded instantaneous cutting force prediction method of any one of the first aspect.

[0049] According to a fifth aspect of the present application, a computer program product is provided, comprising a computer program which, when executed on a computer, causes the computer to perform the physical information embedded instantaneous cutting force prediction method of any one of the first aspect.

[0050] Overall, the above technical solutions conceived by the present application can achieve the following beneficial effects:

[0051] (1) The instantaneous cutting force prediction method provided by the present application inputs the machining time sequence data in the workpiece material machining process into the instantaneous cutting force prediction model to extract the time sequence features, inputs the static machining parameters into the theoretical mechanical cutting force model, calculates the corresponding theoretical cutting force, and splices the theoretical cutting force and the static machining parameters to obtain the mechanism feature vector. The mechanism feature vector is used as the physical information embedded instantaneous cutting force prediction model, and the mechanism feature vector and the time sequence features are spliced in the full connection layer of the model to predict the instantaneous cutting force. The static machining parameters reflect different machining conditions in the workpiece material machining process. The static machining parameters and the corresponding theoretical cutting force constitute the mechanism feature vector, which is input into the model to participate in the prediction of the instantaneous cutting force together with the time sequence features, which can enhance the generalization ability of the model under different working conditions and speed up the calculation speed of the model to improve the real-time performance of the prediction. Moreover, the machining time sequence data including the tool vibration signal and the machine tool motor current signal and the static machining parameters including the speed of the tool, the feed rate and the cutting depth are used to predict the cutting force, so that the present application considers the direct process fluctuation (features of time sequence data) and the long-term force trend (numerical features of theoretical cutting force) in the cutting process at the same time, and improves the prediction accuracy.

[0052] (2) Further, the feature extraction network designed by the present application positions the convolution module after the attention layer, establishing a hierarchical feature representation structure. When the attention layer captures global features, the convolution refines these features into more fine-grained local features. Specifically, based on the global features captured by the attention layer, the time sequence features are further convoluted to capture local dependency relationships and patterns, which is crucial for accurately predicting high-frequency non-stationary time series. This hierarchical feature representation structure significantly enhances the recognition and classification ability of the model.

[0053] (3) As preferred, the attention layer uses a multi-scale attention mechanism to extract time series feature information at different scales. Through the use of multi-scale attention mechanism and convolution module, the multi-scale features of time series signal are fully extracted, and the global features and local features are effectively fused, which significantly improves the representation learning ability of the model, realizes the deep feature extraction from spatial scale to time sequence dimension, and improves the cutting force prediction accuracy.

[0054] (4) The present application provides a specific theoretical cutting force calculation method, which integrates Merchant cutting theory and simplified Johnson-Cook constitutive model to construct a theoretical mechanical cutting force model to calculate the cutting force value. Specifically, Merchant's cutting theory is based on the force balance in the cutting process to establish the relationship between the theoretical cutting force and the shear force on the shear surface; in the cutting process, the shear yield strength τ s The flow stress σ calculated by the simplified Johnson-Cook model without the thermal softening term provides a theoretical basis for the prediction of cutting force in the machining process. The provided theoretical cutting force calculation method comprehensively covers the machining geometry information and static machining parameters involved in the theoretical mechanical cutting force model, which can reduce the number of model training iterations, significantly reduce the training time and convergence speed of the model, and also enhance the interpretability of the model. BRIEF DESCRIPTION OF DRAWINGS

[0055] Figure 1 The physical information embedded in the embodiment of the present application is a schematic diagram of the instantaneous cutting force prediction method.

[0056] Figure 2 The instantaneous cutting force prediction model structure in the embodiment of the present application is a schematic diagram.

[0057] Figure 3 The training prediction result verification diagram based on the boring machining data set in the embodiment of the present application is a schematic diagram.

[0058] Figure 4 The schematic diagram of the time series signal acquisition module provided in the embodiment of the present application is a schematic diagram.

[0059] Figure 5 The schematic diagram of the electronic device provided in the embodiment of the present application is a schematic diagram. DETAILED DESCRIPTION

[0060] In order to make the purpose, technical scheme and advantages of the present application clearer, the present application will be further described in detail below in combination with the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and do not limit the present application. In addition, the technical features involved in each embodiment of the present application described below can be combined with each other as long as they do not conflict with each other.

[0061] Embodiment 1

[0062] As Figure 1 shown, the physical information embedded instantaneous cutting force prediction method provided in the embodiment of the application mainly comprises:

[0063] In the workpiece material processing process, multi-channel processing time sequence data and static processing parameters are collected; in the embodiment of the application, the processing time sequence data includes a tool vibration signal and a machine tool motor current signal, and the static processing parameters include a tool rotation speed, a feed rate f and a cutting depth a p ;

[0064] The processing time sequence data is input into a feature extraction network of a trained instantaneous cutting force prediction model to obtain features of the processing time sequence data;

[0065] The static processing parameters are input into a theoretical mechanical cutting force model to calculate a corresponding theoretical cutting force; and a mechanism feature vector is obtained by splicing the theoretical cutting force and the static processing parameters, and the mechanism feature vector is input into a full connection layer of the trained instantaneous cutting force prediction model;

[0066] In the full connection layer, the above-mentioned processing time sequence data features and the mechanism feature vector are spliced and regressed to obtain an instantaneous cutting force prediction result.

[0067] As a preferred implementation manner, the features of the processing time sequence data include global time sequence features and local time sequence features, and correspondingly, as Figure 2 shown, the above-mentioned instantaneous cutting force prediction model includes a feature extraction network and a full connection layer; wherein the feature extraction network includes an embedding layer, a position encoding layer, an attention layer, a feedforward layer, a normalization layer and a convolution module connected in sequence.

[0068] The embedding layer is used to convert the processing time sequence data (low-dimensional dense data) into a high-dimensional time sequence vector, facilitating subsequent processing of the model.

[0069] The position encoding layer is used to perform position encoding on the high-dimensional time sequence vector to obtain encoded processing time sequence data, facilitating the attention mechanism to learn position information, which is used to reflect the time sequence order in the time sequence data. In the embodiment of the application, a classical sin, cos encoding method in the transformer is used, and the formula is as follows:

[0070]

[0071] Wherein, PE represents position encoding, pos represents the position of an input vector (i.e. token, embedding layer output) to be solved in the entire input vector (transformed high-dimensional vector), the position encoding is a vector (embedding), i represents the index of each element in the vector, dmodel vector dimension representing position coding.

[0072] The attention layer is used to calculate the weight of the features of different scales of the encoded processing time series data by using an attention mechanism, and to obtain fused global time series features by weighted sum of the features of different scales with corresponding weights. In the embodiment of the present application, the attention weight distribution of the Value is realized by the attention convergence of the Query and the Key, and the final output result is generated. Taking a 1 The attention mechanism of the vector and other vectors is taken as an example, a 1 The q vector and the k vector of other input vectors are multiplied respectively and passed through a softmax layer to obtain a 1 The matching value with other input vectors, and the matching value is weighted multiplied with the v vector of each input vector to obtain the output vector. This process can be described by the formula as follows:

[0073]

[0074] Wherein, Q, K, V are the matrix representation of q, k, v of each input vector (encoded processing time series data), d k is a scaling factor, which is used to make the dot product result fall in the area with larger gradient of the softmax function, so as to speed up the training process of the model.

[0075] As a preferred, the attention layer uses a multi-scale attention mechanism to extract time series feature information at different scales.

[0076] The feedforward layer is used to perform nonlinear mapping on the fused global time series features, and after normalization by the normalization layer, the normalized global time series features are obtained. In the embodiment of the present application, the feedforward layer includes a linear layer and a nonlinear activation function, which is used to enhance the ability of the model to learn nonlinear mapping relationship; the normalization layer uses layer normalization technology to improve the convergence speed and stability of the model during training.

[0077] The convolution module is used to extract local features of the normalized global time series features to obtain more fine-grained local time series features; in the embodiment of the present application, the convolution module includes convolution layers, batch normalization layers, activation functions and fully connected layers connected in sequence.

[0078] The fully connected layer of the instantaneous cutting force prediction model is used to splice the normalized global time series features, local time series features and mechanism feature vectors and regress to obtain the instantaneous cutting force prediction result.

[0079] As a preferred implementation, the theoretical mechanical cutting force model in the embodiment of the present application calculates the theoretical cutting force value by integrating Merchant cutting theory, oblique cutting theory and simplified Johnson-Cook constitutive model calculation theory.

[0080] Merchant's cutting theory is based on the force balance in the cutting process, and the relationship between the cutting force and the shear force on the shear surface is established:

[0081]

[0082] In the formula, F c is the tangential cutting force, that is, the theoretical cutting force, φ c is the shear angle, β n is the friction angle, α n is the rake angle of the tool, τ s is the shear yield strength of the processed material, and A s is the shear area.

[0083] The oblique cutting theory is used to calculate the shear angle φ c , the friction angle β n and the rake angle α n of the tool in the formula.

[0084] The shear yield strength τ s is determined based on the simplified Johnson-Cook constitutive model. A simplified Johnson-Cook constitutive model is provided in the embodiments of the present application, which describes the relationship between the flow stress and the strain, strain rate and temperature:

[0085]

[0086] In the formula, σ is the flow stress, ∈ is the shear plastic strain; v is the relative speed between the tool and the workpiece material, which is determined based on the rotational speed of the tool; h is the undeformed chip thickness, T is the temperature of the workpiece material, T room is the room temperature, T melt is the melting point of the workpiece material, A is the yield strength (MPa), B is the strain hardening coefficient (MPa), C is the strain rate sensitivity coefficient, n is the strain hardening index, and m is the thermal softening index.

[0087] The flow stress σ is used instead of the shear yield strength τ s , and the shear area A s can be determined by the cutting depth a p and the feed rate f:

[0088] A s = a p f sin φ c

[0089] The theoretical cutting force calculation formula is obtained:

[0090]

[0091] As a preferred implementation, in the embodiment of the present application, the instantaneous cutting force prediction model is trained by using a data set to obtain the trained instantaneous cutting force prediction model; wherein the samples in the data set include multi-channel machining time series data and static machining parameters, and the label is the cutting force time series data in the machining process, i.e., the cutting force corresponding to the machining time series data and static machining parameters at different time.

[0092] The static machining parameters are input into the theoretical mechanical cutting force model to obtain the corresponding theoretical cutting force value, and after being spliced with the static machining parameters, a mechanism feature vector is formed.

[0093] The machining time series data and the mechanism feature vector are input into the instantaneous cutting force prediction model to obtain the corresponding cutting force prediction result.

[0094] The instantaneous cutting force prediction model is trained by minimizing the difference loss between the cutting force prediction result and the corresponding label.

[0095] In an optional embodiment, data collection is performed in a vertical boring machining scene, the tool used is a triangular boring tool, and the workpiece used is a titanium alloy test piece. The vibration and current signals are collected by a sensor, and then the AD conversion module is used to finally obtain high-quality boring machining time series data and static parameters. All signals are collected at a frequency of 10 kHz for 50 s. The data set is randomly divided into a training set, a validation set and a test set, which respectively account for 80%, 10% and 10% of the data set. The model is initialized with random weights using the xavier initialization technique, and the model is optimized using the Adam optimizer.

[0096] After training, the performance of the model is evaluated using evaluation indexes such as MAE, MSE and R2. The trailing face wear bandwidth detection model reaches 0.0052, 0.0126 and 0.9578 in the MAE, RMSE and R2 evaluation indexes, i.e., the prediction accuracy of the method of the present application is high. As shown in Figure 3 the predicted value is basically consistent with the true value, which also indicates that the prediction accuracy of the method of the present application is high.

[0097] The instant cutting force prediction method provided by the application, by inputting the machining time sequence data in the workpiece material machining process into the instant cutting force prediction model for time sequence feature extraction, inputting the static machining parameters into the theoretical mechanical cutting force model, calculating the corresponding theoretical cutting force, splicing the theoretical cutting force and the static machining parameters to obtain the mechanism feature vector, the mechanism feature vector is embedded into the instant cutting force prediction model as physical information, and the mechanism feature vector and the time sequence feature are spliced in the full connection layer of the model to predict the instant cutting force. The static machining parameters reflect different machining conditions in the workpiece material machining process, the static machining parameters and the corresponding theoretical cutting force constitute the mechanism feature vector input into the model, and the time sequence feature participates in the instant cutting force prediction, which can enhance the generalization ability of the model under different conditions, and speed up the calculation speed of the model, to improve the real-time of prediction. And, the machining time sequence data including tool vibration signal and machine tool motor current signal and the static machining parameters including speed of the tool, feed rate f and cutting depth a p The application simultaneously considers the direct process fluctuation (time sequence data features) and long-term force trend (theoretical cutting force numerical features) in the cutting process, and improves the prediction accuracy.

[0098] Further, the feature extraction network designed by the application positions the convolution module after the attention layer, and establishes a hierarchical feature representation structure. When the attention layer captures global features, the convolution refines these features into more fine-grained local features. Specifically, based on the global features captured by the attention layer, the time sequence features are further convoluted to capture local dependency relationships and patterns, which is crucial for accurately predicting high-frequency non-stationary time sequences. This hierarchical feature representation structure significantly enhances the recognition and classification ability of the model.

[0099] The attention layer uses a multi-scale attention mechanism to extract time sequence feature information at different scales. Through the use of the multi-scale attention mechanism and the convolution module, the multi-scale features of the time sequence signal are fully extracted, and the global features and local features are effectively fused, significantly improving the representation learning ability of the model, realizing deep feature extraction from spatial scale to time sequence dimension, and improving the cutting force prediction accuracy.

[0100] Further, considering the black box characteristics of current deep learning, the calculation process is difficult to understand and has no significant physical meaning, the application embeds the features reflecting the physical information (static machining parameters and theoretical cutting force) of the cutting process into the deep learning model, which brings a certain explainability to the model.

[0101] The application provides a specific theoretical cutting force calculation mode, which integrates Merchant cutting theory, oblique cutting theory and a simplified Johnson-Cook constitutive model to calculate the cutting force value, and has the advantages of solid theoretical basis, clear physical meaning, high calculation efficiency, good versatility and strong engineering practicability compared with the prior art. s The flow stress σ calculated by the simplified Johnson-Cook model without a thermal softening term is used to represent the prediction of the cutting force in the machining process, and provides a theoretical basis for the prediction of the cutting force in the machining process. The theoretical cutting force calculation mode comprehensively covers the machining geometric information and static machining parameters involved in the theoretical mechanical cutting force model, can reduce the number of model training iterations, significantly reduce the training time and convergence speed of the model, and enhance the interpretability of the model.

[0102] Embodiment 2

[0103] The embodiment of the application provides a physical information embedded instantaneous cutting force prediction system, which comprises:

[0104] The trained instantaneous cutting force prediction model and the mechanism feature vector acquisition module; the instantaneous cutting force prediction model comprises a feature extraction network and a full connection layer;

[0105] The feature extraction network is used for extracting the features of the machining time series data; wherein the machining time series data comprises a tool vibration signal and a machine tool motor current signal;

[0106] The mechanism feature vector acquisition module is used for inputting the static machining parameters into the theoretical mechanical cutting force model to calculate the corresponding theoretical cutting force; and the theoretical cutting force and the static machining parameters are spliced to obtain the mechanism feature vector; wherein the static machining parameters comprise the rotational speed of the tool, the feed rate and the cutting depth;

[0107] The full connection layer is used for splicing the mechanism feature vector and the obtained machining time series data features and regression to obtain the instantaneous cutting force prediction result.

[0108] The related technical solutions are described in Embodiment 1, which will not be repeated here.

[0109] Further, the data acquisition module is further included, which is used for acquiring the machining time series data and the machining static parameters in the machining process; wherein the machining time series data comprises the tool vibration signal and the machine tool motor current signal, and the machining static parameters comprise the rotational speed of the tool, the feed rate f and the cutting depth a p .

[0110] Specifically, as Figure 4As shown, the timing signal acquisition module includes sensors, a main control module, and an analog-to-digital converter. The sensors include an accelerometer and a current sensor. The accelerometer is mounted on the tool handle to collect vibration signals, while the current sensor is mounted on the machine tool motor to collect current signals. The main control module sends start / stop signals to each sensor to ensure the synchronization of collected data. The analog-to-digital converter converts the analog signals sent back by the sensors into digital signals for easy computer reading and storage.

[0111] Example 3

[0112] An embodiment of the present invention provides an electronic device, which includes a computer-readable storage medium and a processor; the computer-readable storage medium is used to store executable instructions; the processor is used to read the executable instructions stored in the computer-readable storage medium to execute the instantaneous cutting force prediction method embedded with physical information in Example 1 to obtain an instantaneous cutting force prediction result.

[0113] The electronic device may be a computing device such as a desktop computer, a notebook, a PDA, or a cloud server. The processor may be a central processing unit (CPU), or other general-purpose processors, 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, discrete hardware components, etc. The memory may be used to store computer programs and / or modules, and the processor may perform various functions of the electronic device by running or executing the computer programs and / or modules stored in the memory, and calling the data stored in the memory.

[0114] The relevant technical solutions are described in Example 1 and will not be repeated here.

[0115] Specifically, if Figure 5 As shown, as a specific implementation method, the collected processing timing data and processing static parameters are input into a computer, and the processor in the computer reads the executable instructions stored in the computer-readable storage medium to execute the instantaneous cutting force prediction method embedded with physical information in Example 1 to obtain the instantaneous cutting force prediction result.

[0116] Example 4

[0117] The embodiment of the present application provides a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to realize the steps of the method in the above embodiment 1.

[0118] Specifically, the memory can include a high-speed random access memory, and can further include a non-volatile memory, for example, a hard disk, a memory, a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, at least one disk storage device, a flash memory device, or other volatile solid-state memory devices.

[0119] The related technical solutions are the same as above, and will not be repeated here.

[0120] Embodiment 5

[0121] The embodiment of the present application provides a computer program product, which comprises computer instructions stored in a computer readable storage medium. A processor of a computer device reads the computer instructions from the computer readable storage medium, and the processor executes the computer instructions, so that the computer device executes the steps of the method in the above embodiment 1.

[0122] The related technical solutions are the same as above, and will not be repeated here.

[0123] Those skilled in the art can easily understand that the above description is only the preferred embodiment of the present application, and is not used to limit the present application, and any modification, equivalent replacement and improvement made within the spirit and principle of the present application should be included in the protection scope of the present application.

Claims

1. A physical information embedded instantaneous cutting force prediction method, characterized in that: include: Inputting the processing time series data into the feature extraction network of the trained instantaneous cutting force prediction model to obtain the features of the processing time series data; wherein the processing time series data includes the tool vibration signal and the machine tool motor current signal; Inputting static machining parameters into a theoretical mechanical cutting force model to calculate the corresponding theoretical cutting force; concatenating the theoretical cutting force with the static machining parameters to obtain a mechanism characteristic vector, which is then input into a fully connected layer of the instantaneous cutting force prediction model; wherein the static machining parameters include tool speed, feed rate, and cutting depth; In the fully connected layer, the mechanism feature vector is concatenated with the acquired processing time series data features and regressed to obtain the instantaneous cutting force prediction result; The theoretical mechanical cutting force model is: in, is the theoretical cutting force, is the yield strength, is the strain hardening coefficient, is the strain rate sensitivity coefficient, is the strain hardening exponent, is the thermal softening index, is the shear plastic strain; is the relative speed between the tool and the workpiece material, determined based on the rotational speed of the tool; is the thickness of the undeformed chip, is the temperature of the workpiece material, is room temperature, is the melting point of the workpiece material; is the shear angle, is the friction angle, is the rake angle of the tool, is the cutting depth, is the feed rate.

2. The instantaneous cutting force prediction method based on physical information embedding according to claim 1 is characterized in that: The features of the processed time series data include global time series features and local time series features; The feature extraction network includes: An embedding layer, configured to convert the processed time series data into a high-dimensional time series vector; A position encoding layer, configured to perform position encoding on the high-dimensional time series vector to obtain encoded processed time series data; An attention layer, configured to calculate the weights of features of different scales of the encoded processed time series data using an attention mechanism, and perform weighted summation of the features of different scales using the corresponding weights to obtain a fused global time series feature; A feedforward layer, configured to perform nonlinear mapping on the global temporal features; Normalization layer, used to normalize the global temporal features after nonlinear mapping; The convolution module is used to extract local features from the normalized global time series features to obtain finer-grained local time series features; Correspondingly, the fully connected layer is used to splice and regress the normalized global time series features, the local time series features and the mechanism feature vector to obtain the instantaneous cutting force prediction result.

3. The instantaneous cutting force prediction method based on physical information embedding according to claim 1 is characterized in that: The theoretical mechanical cutting force model is constructed as follows: Establishing theoretical cutting force based on Merchant's cutting theory Relationship with shear force on shear plane: Where, is the shear yield strength of the processed material, is the shear area; The flow stress determined based on the simplified Johnson-Cook constitutive model Instead of the shear yield strength ; wherein the flow stress for: The shear area By the cutting depth and the feed rate Sure: The shear yield strength and the shear area The calculation formula of the theoretical cutting force is substituted into The theoretical mechanical cutting force model is obtained from the relationship between θ and the shear force on the shear surface.

4. The instantaneous cutting force prediction method based on physical information embedding according to claim 2 is characterized in that: The attention mechanism is a multi-scale attention mechanism.

5. A physical information embedded instantaneous cutting force prediction system, characterized in that: include: A trained instantaneous cutting force prediction model and a mechanism feature vector acquisition module; the instantaneous cutting force prediction model includes a feature extraction network and a fully connected layer; The feature extraction network is used to extract features of processing time series data; wherein the processing time series data includes tool vibration signals and machine tool motor current signals; The mechanism characteristic vector acquisition module is used to input static machining parameters into a theoretical mechanical cutting force model to calculate the corresponding theoretical cutting force; and to obtain the mechanism characteristic vector by combining the theoretical cutting force with the static machining parameters; wherein the static machining parameters include the tool rotation speed, feed rate and cutting depth; The fully connected layer is used to splice and regress the mechanism feature vector and the acquired processing time series data features to obtain the instantaneous cutting force prediction result; The theoretical mechanical cutting force model is: in, is the theoretical cutting force, is the yield strength, is the strain hardening coefficient, is the strain rate sensitivity coefficient, is the strain hardening exponent, is the thermal softening index, is the shear plastic strain; is the relative speed between the tool and the workpiece material, determined based on the rotational speed of the tool; is the thickness of the undeformed chip, is the temperature of the workpiece material, is room temperature, is the melting point of the workpiece material; is the shear angle, is the friction angle, is the rake angle of the tool, is the cutting depth, is the feed rate.

6. The instantaneous cutting force prediction system with physical information embedding according to claim 5 is characterized in that: The features of the processed time series data include global time series features and local time series features; The feature extraction network includes: An embedding layer, configured to convert the processed time series data into a high-dimensional time series vector; A position encoding layer, configured to perform position encoding on the high-dimensional time series vector to obtain encoded processed time series data; An attention layer, configured to calculate the weights of features of different scales of the encoded processed time series data using an attention mechanism, and perform weighted summation of the features of different scales using the corresponding weights to obtain a fused global time series feature; A feedforward layer, configured to perform nonlinear mapping on the global temporal features; Normalization layer, used to normalize the global temporal features after nonlinear mapping; The convolution module is used to extract local features from the normalized global time series features to obtain finer-grained local time series features; Correspondingly, the fully connected layer is used to splice and regress the normalized global time series features, the local time series features and the mechanism feature vector to obtain the instantaneous cutting force prediction result.

7. An electronic device, characterized in that: comprising a computer-readable storage medium and a processor; The computer-readable storage medium is used to store executable instructions; The processor is configured to read the executable instructions stored in the computer-readable storage medium to execute the instantaneous cutting force prediction method embedded with physical information as described in any one of claims 1 to 4.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the instantaneous cutting force prediction method embedded with physical information as described in any one of claims 1 to 4 is implemented.

9. A computer program product, characterized in that The invention comprises a computer program, which, when running on a computer, enables the computer to execute the instantaneous cutting force prediction method embedded with physical information as claimed in any one of claims 1 to 4.

Images