Tool wear real-time prediction method fusing wear and jumping mechanism-three-dimensional vibration data

Abstract

The invention discloses a tool wear real-time prediction method fusing wear and jumping mechanism-three-dimensional vibration data, and belongs to the field of machine tool state monitoring and intelligent manufacturing. According to the method, a three-dimensional acceleration sensor is arranged near a main shaft or a cutter handle, three-dimensional vibration signals in the machining process are obtained, and a cutting force mechanism model is established in combination with cutting parameters and geometric features of the cutter; and tool wear correction and bounce disturbance items are introduced into the mechanism model, and prediction calculation of the three-way cutting force is carried out. And then, a cutting force prediction result and vibration signal features are fused, a comprehensive feature vector is constructed, and real-time prediction of the tool abrasion loss is realized based on a deep learning regression model. According to the method, physical rule constraints of mechanism modeling and data driving advantages of vibration signals are fully utilized, the accuracy and real-time performance of tool wear prediction are effectively improved, and important engineering application value is provided for improving the machining quality, prolonging the service life of a tool and achieving intelligent production.

Description

Technical Field

[0001] This invention belongs to the field of CNC cutting and machining, and relates to a method for real-time prediction of tool wear that integrates wear and vibration mechanism—three-dimensional vibration data. Background Technology

[0002] As a crucial consumable in CNC machine tool manufacturing, tooling wear significantly impacts machining accuracy and surface quality, leading to increased processing time and costs. When tool wear reaches 10%, processing time increases by 40%, while costs rise by 30%. 7-20% of total downtime is caused by unexpected downtime due to tool wear and abnormal breakage. However, tool wear prediction can significantly reduce tool consumption. By predicting tool wear and replacing them promptly, tool life can be extended by over 40%. Furthermore, severe tool wear poses a threat to personnel and equipment safety; manufacturing accidents caused by tool damage and breakage account for over 10% of all accidents.

[0003] In existing technologies, tool wear prediction mainly employs data-driven, fault mechanism-based, and knowledge-learning-based methods. However, these methods still cannot fully cover the complex and variable working conditions in actual engineering. Data-driven methods extract effective information from massive amounts of collected signals and build predictive models to monitor tool wear in real time. Although this method does not rely excessively on expert experience, given the explosive growth in sensor monitoring data collection, it still faces challenges such as difficulty in acquiring signal data and weak generalization ability of monitoring models for tools of different models and operating conditions. Furthermore, data-driven monitoring models often fail to consider prior knowledge such as tool structural parameters and wear mechanisms, making it difficult to accurately predict tool wear in complex scenarios. Fault mechanism models, on the other hand, provide a deeper analysis of the physical interpretation and meaning of model parameters, describe the tool wear mechanism and characteristics, and have clear physical meanings for process variables, effectively complementing data-driven methods. Therefore, combining mechanism models and data models can fully utilize the advantages of each model, improving prediction accuracy, real-time monitoring, and reliability.

[0004] In CNC milling, the tool wear and degradation process is complex, and the machining system suffers from manufacturing errors of the machine tool, geometric errors of the tool and fixture, and installation errors of the tool holder-spindle joint. Tool runout is unavoidable during machining. Due to tool wear and runout, the proportion of cases where the instantaneous cut thickness (IUCT) is less than the minimum cut thickness is high. Elastic deformation occurs on the workpiece surface, resulting in some surface material not being removed, thus increasing the surface morphology error of the machined material. Therefore, establishing a real-time tool wear prediction method that considers tool wear and runout is crucial.

[0005] Three-dimensional vibration signals can comprehensively reflect the dynamic response of the cutting tool during the cutting process. Their amplitude and frequency characteristics change significantly with changes in tool wear and runout, exhibiting a clear physical correlation. Compared to methods relying on force gauges to directly acquire cutting force signals, vibration signal acquisition is more convenient and does not cause additional interference to the machine tool's machining process. By placing triaxial accelerometers near the spindle box or tool holder, high-frequency vibration data can be acquired in real time, enabling tool condition monitoring without altering the machine tool structure. As tool wear deepens, the vibration energy and characteristic distribution generated during cutting change, providing observable evidence for establishing tool wear prediction models. Therefore, real-time prediction of tool wear based on three-dimensional vibration signals has strong feasibility and application value.

[0006] Chinese invention patent CN 114102260 B discloses a mechanism-data fusion-driven method for monitoring tool wear under varying operating conditions. This method reconstructs the cutting force using spindle acceleration and motor current signals, thereby replacing a force gauge to monitor tool wear and predict remaining tool life. While this method overcomes the limitations of fixed cutting conditions to some extent, its signal processing is relatively simple, relying solely on labels for differentiation and failing to fully exploit the dynamic characteristics inherent in the signals. Furthermore, the low sampling frequency of the spindle current signal makes it difficult to effectively characterize the dynamic features of the tool under high-speed cutting conditions. Therefore, existing methods still have limitations in describing wear mechanisms and utilizing high-frequency signals under complex operating conditions.

[0007] Chinese invention patent CN 117094210 A discloses a tool wear prediction method based on mechanism-data fusion. This method utilizes a power signal-based mechanism prediction model to obtain theoretical predictions of tool wear, combines these predictions with sensor signal features, introduces a physical consistency loss function, and employs a gated recurrent unit (GRU) network to address the temporal nature of tool wear, thus improving prediction accuracy and model interpretability to some extent. However, this method fails to fully consider the impact of tool wear and vibration mechanisms on cutting force characteristics, and also fails to fully utilize the dynamic characteristics inherent in three-dimensional vibration signals. Therefore, its adaptability and reliability under complex cutting conditions still have room for improvement.

[0008] In summary, existing tool wear prediction methods suffer from the following problems: the mechanistic model does not fully consider the actual machine tool structure and operating conditions; the machining process relies on a force gauge to obtain cutting forces; and high-frequency signal characteristics are not adequately utilized, leading to difficulties in guaranteeing prediction accuracy. There is an urgent need to develop new tool wear prediction methods to address these issues. Summary of the Invention

[0009] The purpose of this invention is to overcome the shortcomings of the prior art and provide a real-time prediction method for tool wear that integrates wear and vibration mechanism—three-dimensional vibration data.

[0010] To achieve the above objectives, the present invention employs the following technical solution:

[0011] A real-time tool wear prediction method integrating wear and vibration mechanism—three-dimensional vibration data, comprising the following steps:

[0012] Step 1) Arrange a three-dimensional accelerometer near the machine tool spindle or tool holder to collect three-dimensional vibration signals during the machining process, and simultaneously obtain cutting parameters and tool geometry features;

[0013] Step 2) Using the collected cutting parameters and tool geometry, establish an instantaneous invariant cutting thickness model and a basic cutting force mechanism model using the infinitesimal element method;

[0014] Step 3) Based on the aforementioned mechanism model and cutting force test data, an iterative optimization method is used to identify tool runout parameters, thereby obtaining the tool radial runout and phase information;

[0015] Step 4) Introduce the flank wear correction and the identified runout disturbance term into the mechanism model, perform real-time cutting force calculation, and obtain the triaxial cutting force prediction result that simultaneously considers the effects of tool wear and runout.

[0016] Step 5) Extract time-domain, frequency-domain, and time-frequency-domain features from the acquired three-dimensional vibration signal, and fuse them with the cutting force features output by the mechanism model to construct a comprehensive feature vector for prediction;

[0017] Step 6) Establish a regression model based on deep learning methods to form a tool wear prediction model and realize real-time prediction of tool wear.

[0018] Furthermore, the cutting parameters in step 1 include spindle speed, feed rate, depth of cut, and width of cut; the tool geometry includes tool diameter, number of cutting edges, helix angle, and rake angle and clearance angle.

[0019] Furthermore, the specific method for establishing the instantaneous invariant cutting thickness model and the basic cutting force mechanism model using the infinitesimal element method in step 2 is as follows:

[0020] First, using the collected cutting parameters and tool geometry, the tool cutting edge is discretized into several infinitesimal elements along the rotation angle using the infinitesimal element method, and an instantaneous constant cutting thickness calculation model is established to obtain the expression for h(t).

[0021] Secondly, by substituting the instantaneous constant cutting thickness h(t) into the cutting force mechanism model, a basic cutting force calculation formula is constructed:

[0022] F(t) = K t ·h (t) ·a p

[0023] Where F(t) is the cutting force, K t h is the cutting force coefficient. (t) For the instantaneous constant cutting thickness, a p This represents the cutting depth.

[0024] Furthermore, the method for identifying tool runout parameters in step 3 is as follows:

[0025] First, in the initial cutting stage, based on the milling force test data in the X and Y directions and the radial wear value collected by the force measuring instrument, the position angle of one revolution of the tool is selected as the sample point;

[0026] Secondly, the initial values ​​of the tool radial runout and phase are set and substituted into the instantaneous cutting thickness model and cutting force model to calculate the theoretical cutting force value, which is then compared with the experimentally measured cutting force to form the sum of squared errors.

[0027] Then, the iteration step size and runout parameter constraint range are set, and the parameter values ​​are continuously updated through iterative optimization to gradually reduce the difference between the theoretical and measured cutting forces.

[0028] Finally, when the sum of squared errors reaches its minimum value, the corresponding radial runout and phase of the tool are the optimal identification results.

[0029] Furthermore, the specific method for introducing the flank wear correction and the identified runout disturbance term into the mechanism model in step 4 is as follows:

[0030] First, tool wear correction is introduced into the cutting force mechanism model constructed in step 2, taking into account the flank wear amount V. b The effect on effective cutting geometry increases Δh(V) b Correction of instantaneous cutting thickness:

[0031] h (t)1 =h (t) +Δh(V b )

[0032] Wherein, Δh(V b () represents the wear correction amount.

[0033] Then, the radial runout of the tool is introduced into the model, with the runout amount being Δr and the phase being φ, to correct the instantaneous cutting thickness h. (t)2 for:

[0034] h (t)2 =h (t)1 +Δr·cos(wt+φ)

[0035] Where w is the principal axis angular velocity.

[0036] Finally, using the corrected instantaneous cutting thickness h (t)2 A cutting force prediction mechanism model that simultaneously considers tool wear and runout factors is constructed.

[0037] Furthermore, the specific operation of the mechanism model in step 4 to calculate the cutting force in real time during the machining process is as follows:

[0038] First, obtain the instantaneous cutting thickness h after considering tool wear and runout correction. (t)2 Then, the values ​​are substituted into the milling force prediction mechanism model to calculate the three-dimensional cutting forces in real time during the machining process.

[0039] Secondly, the forces acting on the discrete elements of the cutting edge are integrated using the infinitesimal element method to obtain the tangential force, radial force, and axial force components:

[0040]

[0041] in, , , Indicates the tangential, radial, and axial cutting force components. , , It is expressed as the cutting force coefficient.

[0042] Finally, the force results of all cutting edges are superimposed to obtain the cutting force prediction data that takes into account the effects of tool wear and runout.

[0043] Furthermore, the method for constructing the comprehensive feature vector in step 5 is as follows:

[0044] First, the collected three-dimensional vibration signal and the triaxial cutting force signal calculated by the mechanism model are preprocessed, and then their feature values ​​are extracted from the time domain, frequency domain and time-frequency domain respectively.

[0045] Then, using correlation analysis, features highly correlated with tool wear were selected from the feature values;

[0046] Finally, the selected vibration signal features are fused with the cutting force features to construct a comprehensive feature vector, which serves as the input to the tool wear prediction model.

[0047] Furthermore, in step 6, a regression model is established based on deep learning methods, preferably using a neural network structure with the ability to model temporal features.

[0048] Compared with the prior art, the present invention has the following beneficial effects:

[0049] This invention proposes a real-time tool wear prediction method that integrates wear and runout mechanisms—three-dimensional vibration data. In the cutting force prediction modeling process, it simultaneously considers two key factors: tool wear and runout. First, a milling force mechanism model is established using the micro-element method. The cutting edge of the tool is discretized into several micro-element units along the tool rotation angle, and the force on each micro-element is integrated to obtain the overall cutting force expression. Based on this, the influence of flank wear on tool geometry parameters is introduced, and the instantaneous constant cutting thickness is corrected, enabling the model to reflect the cutting force variation trend caused by tool wear. Furthermore, a radial runout disturbance term is added to the instantaneous constant cutting thickness to construct a cutting force correction model that reflects the periodic fluctuation characteristics of tool runout, thus obtaining a mechanism prediction model that simultaneously includes wear and runout effects. Unlike existing modeling methods that only consider wear or ignore runout factors, this invention simultaneously introduces wear correction and runout correction in cutting force prediction. This not only more accurately characterizes the actual force state of the tool but also effectively avoids the problems of unclear physical meaning and limited applicability of prediction results caused by single empirical correction methods. The mechanistic model of this invention has clear physical consistency and can realistically reflect the coupling effect of tool wear and runout on cutting force, making the prediction results more adaptable to complex working conditions. The improved mechanistic modeling method of this invention can predict cutting force changes using machining parameters and tool geometry without relying on expensive force gauges for real-time monitoring. This provides a reliable data foundation and theoretical support for subsequent tool wear prediction and has strong engineering application value.

[0050] This invention combines a cutting force prediction mechanism model with three-dimensional vibration signal features to construct a tool wear prediction model that integrates mechanistic and dynamic signal features. It fully leverages the complementary advantages of mechanistic modeling and data-driven methods, enabling high-precision and high-reliability tool wear prediction under complex working conditions. Compared to existing methods that rely solely on a single signal or model, this invention offers the following significant advantages: First, it utilizes a cutting force mechanism model that considers tool wear and runout to output prediction data with clear physical meaning, enhancing the interpretability and reliability of the prediction process. Second, through three-dimensional vibration signal feature extraction, it can sensitively capture the dynamic characteristics of the interaction between the tool and the workpiece, compensating for the inadequacy of a single mechanism-driven approach in fully reflecting the implicit characteristics of the tool wear process. Third, it fuses the mechanistic model output with vibration features, forming multi-source complementary feature inputs, improving the robustness and generalization ability of the wear prediction model under different cutting conditions. Fourth, it employs a deep learning regression method for wear prediction, possessing real-time online processing capabilities, ensuring the applicability of the method in actual machining processes.

[0051] In summary, this invention balances the physical rationality of the mechanistic model with the sensitivity of sensor data, improving the accuracy and reliability of tool wear prediction, and has strong engineering application value and promotion potential. Attached Figure Description

[0052] Figure 1 This is a flowchart illustrating the overall process of the method of the present invention.

[0053] Figure 2 This is a flowchart illustrating the tool runout parameter identification process of the present invention.

[0054] Figure 3 This is a flowchart of the milling force calculation for the present invention. Detailed Implementation

[0055] The present invention will now be described in further detail with reference to the accompanying drawings:

[0056] Example 1

[0057] A real-time tool wear prediction method integrating wear and vibration mechanism-three-dimensional vibration data, such as Figure 1 As shown, it includes the following steps:

[0058] Step 1) Collect data such as three-dimensional vibration signals, cutting parameters, and tool geometry during the machining process;

[0059] Step 2) Using the collected cutting parameters and tool geometry, establish an instantaneous invariant cutting thickness model and a basic cutting force mechanism model using the infinitesimal element method;

[0060] Step 3) Based on the aforementioned mechanism model and cutting force test data, an iterative optimization method is used to identify tool runout parameters, thereby obtaining the tool radial runout and phase information;

[0061] Step 4) Introduce the flank wear correction and the identified runout disturbance term into the mechanism model, perform real-time cutting force calculation, and obtain the triaxial cutting force prediction result that simultaneously considers the effects of tool wear and runout.

[0062] Step 5) Extract time-domain, frequency-domain, and time-frequency-domain features from the acquired three-dimensional vibration signal, and fuse them with the cutting force features output by the mechanism model to construct a comprehensive feature vector for prediction;

[0063] Step 6) Establish a regression model based on deep learning methods, and use the comprehensive feature vector as input to realize real-time prediction of tool wear.

[0064] Example 2

[0065] Combination Figure 2 and Figure 3By using an accelerometer installed near the spindle or tool holder of a machine tool, three-dimensional vibration signals during the machining process are collected. Combined with cutting parameters obtained from the CNC system and tool geometry, an instantaneous cutting thickness and cutting force mechanism model is established. Based on this, an iterative method is used to identify tool runout parameters, and the identification results, along with a tool wear correction term, are incorporated into the mechanism model to achieve cutting force prediction considering the effects of wear and runout. Furthermore, the cutting force prediction results are fused with vibration signal features to construct a comprehensive feature vector. Finally, a deep learning regression model is used to achieve real-time prediction of tool wear. The key steps are as follows:

[0066] Step (1) Data Acquisition: Install a three-dimensional accelerometer near the tool holder on the machine tool spindle to synchronously acquire three-dimensional vibration signals (a) during the machining process. x (t), a y (t), a z (t)). Simultaneously, cutting parameters (including spindle speed n, feed rate v) are obtained from the CNC system. f Feed per tooth f z Cutting depth a p Cutting width a e The cutting parameters and geometric features (diameter D, number of cutting edges Z, helix angle λ, rake angle γ, clearance angle α) are used as inputs for the mechanistic model to calculate the instantaneous cutting thickness and cutting force.

[0067] Step (2) Establishing the basic mechanism model: Based on the cutting parameters and tool geometry obtained in step 1, the instantaneous invariant cutting thickness model and the basic cutting force mechanism model are established using the infinitesimal element method. The specific process is as follows:

[0068] 1) Modeling of instantaneous constant cutting thickness

[0069] The cutting edge of the tool is discretized into several infinitesimal elements along the tool rotation angle θ. Assume the feed per tooth is f. z If the helix angle is λ, then the instantaneous constant cutting thickness of a certain cutting edge at position angle θ can be expressed as:

[0070] (1)

[0071] in, For the j-th blade at position angle The instantaneous cutting thickness at that point remained unchanged.

[0072] 2) Micro-element modeling of cutting force: The micro-element cutting force in the tangential, radial, and axial directions can be expressed as:

[0073] (2)

[0074] in, , , Expressed as the cutting force coefficient, Let be the infinitesimal length of the cutting edge.

[0075] 3) Basic cutting force calculation: For the j-th cutting edge in the effective cutting angle domain Ω j Integrating the infinitesimal force within the teeth and superimposing it on all teeth yields the overall three-dimensional cutting force:

[0076] (3)

[0077] Among them, Ω j Let Z be the cutting angle domain of the j-th cutting edge, and Z be the number of tool teeth.

[0078] The basic cutting force mechanism model obtained by the above modeling takes cutting parameters and tool geometry as inputs and outputs uncorrected triaxial basic cutting forces.

[0079] Step (3) Tool runout parameter identification: such as Figure 2 As shown, tool runout is a static characteristic caused by the spindle-tool assembly and can be identified during the initial cutting stage. The identification results can be directly applied to subsequent machining processes. To ensure the accuracy of the cutting force prediction model, this step uses an iterative optimization method to identify the tool radial runout and phase. The specific process is as follows:

[0080] 1) Based on the sampling frequency of the force measuring instrument, select the position angle of one revolution of each milling cutter as the sample point, and obtain the test values ​​of the milling force in the X and Y directions and the radial wear value of the tool.

[0081] 2) Set the initial values ​​for tool runout parameters and radial wear values ​​( , , The initial values ​​are substituted into the instantaneous cutting thickness model, and then the milling force coefficient is obtained by substituting the experimentally measured milling force into the force model. Finally, the sum of squared milling force errors at various positions of the milling cutter is calculated. .

[0082] 3) Set the iteration step size , and the maximum value of the bounce parameter , The iterative algorithm is implemented iteratively. The radial wear value of the tool is substituted into the instantaneous milling thickness model to calculate the sum of the squared differences between the measured and theoretical milling forces at various position angles with tool runout and radial wear. .

[0083] 4) Selecting the iterative process to make When the minimum value is obtained and This represents the optimal value for the tool runout parameter.

[0084] Step 4: Cutting force prediction: (e.g.) Figure 3 As shown, based on the basic mechanism model in step 2 and the runout parameters obtained in step 3, tool wear correction is introduced to establish a cutting force prediction model.

[0085] 1) Tool wear mainly occurs on the flank face, and increases with the amount of wear V b As the radius of the cutting edge increases, the effective radial dimension of the cutting edge decreases, leading to an adjustment in the cutting force coefficient. Let the reduction in the tool radial dimension be... The cutting force coefficient can be expressed as:

[0086] (4)

[0087] In the formula, , , , , , The parameters are constants of the milling force coefficient in a specific machining process. They are constants determined by the machining parameters, the material properties of the workpiece, and the tool type. They can be determined by fitting specific milling experiments.

[0088] 2) Combining wear correction and runout disturbance, the instantaneous cutting thickness is updated to obtain the corrected thickness h. j (φ,t).

[0089] 3) By superimposing all cutting edges, a cutting force prediction model considering tool wear and runout is obtained:

[0090]

[0091] in, It refers to the number of cutting teeth. It is the total number of discrete units.

[0092] Step (5) Feature Extraction and Fusion: During the machining process, the collected three-dimensional vibration signal and the three-dimensional cutting force signal calculated in step 4 are preprocessed, and corresponding feature parameters are extracted from the time domain, frequency domain, and time-frequency domain. Subsequently, the correlation analysis method is used to screen out features that are highly correlated with the tool wear, and the screened vibration signal features are fused with the cutting force features to form a comprehensive feature vector, which is used as the input for the subsequent tool wear prediction model.

[0093] Step (6) Tool Wear Prediction: Based on the comprehensive feature vector obtained in step (5), a regression model is established using deep learning to achieve real-time prediction of tool wear. The regression model preferably uses a neural network structure with time-series feature modeling capabilities to capture the dynamic evolution of tool wear with machining time and cutting conditions, thereby improving prediction accuracy and real-time performance.

[0094] In summary, the method proposed in this invention, based on a cutting force prediction mechanism model and combined with three-dimensional vibration signals, achieves a fusion prediction driven by both mechanism and data. By placing accelerometers near the machine tool spindle or tool holder, three-dimensional vibration signals during the machining process can be collected in real time and combined with cutting parameters and tool geometry to form a complete data input. By introducing tool wear correction and runout disturbance into the mechanism model and using iterative algorithms to identify tool runout parameters, a three-dimensional cutting force prediction result that accurately reflects the actual working conditions is obtained. Subsequently, the cutting force prediction features are fused with vibration signal features to construct a comprehensive feature vector, and a regression model is established based on deep learning methods to achieve real-time prediction of tool wear. This method avoids the limitations of relying entirely on force gauges, fully utilizes the complementarity between vibration signals and mechanism calculation results, and improves the accuracy and applicability of the prediction model. Through dynamic updates of the prediction results, a reliable basis can be provided for tool management and maintenance in intelligent production environments, which has significant engineering application value for improving machining quality, extending tool life, and realizing intelligent machining.

[0095] The above content is only for illustrating the technical concept of the present invention and should not be construed as limiting the scope of protection of the present invention. Any modifications made to the technical solution based on the technical concept proposed in this invention shall fall within the scope of protection of this invention.

Claims

1. A method for real-time prediction of tool wear integrating wear and vibration mechanism-three-dimensional vibration data, characterized in that, Includes the following steps: Step 1) Arrange a three-dimensional accelerometer near the machine tool spindle or tool holder to collect three-dimensional vibration signals during the machining process, and simultaneously obtain cutting parameters and tool geometry features; Step 2) Using the collected cutting parameters and tool geometry, establish an instantaneous invariant cutting thickness model and a basic cutting force mechanism model using the infinitesimal element method; Step 3) Based on the aforementioned mechanism model and cutting force test data, an iterative optimization method is used to identify tool runout parameters, thereby obtaining the tool radial runout and phase information; Step 4) Introduce the flank wear correction and the identified runout disturbance term into the mechanism model, perform real-time cutting force calculation, and obtain the triaxial cutting force prediction result that simultaneously considers the effects of tool wear and runout. Step 5) Extract time-domain, frequency-domain, and time-frequency-domain features from the acquired three-dimensional vibration signal, and fuse them with the cutting force features output by the mechanism model to construct a comprehensive feature vector for prediction; Step 6) Establish a regression model based on deep learning methods to form a tool wear prediction model and realize real-time prediction of tool wear.

2. The method for real-time prediction of tool wear by fusing wear and vibration mechanism-three-dimensional vibration data according to claim 1, characterized in that, In step 1), the obtained cutting parameters include spindle speed, feed rate, depth of cut, and width of cut; the tool geometry features include tool diameter, number of teeth, helix angle, rake angle, and clearance angle.

3. The method for real-time prediction of tool wear by fusing wear and vibration mechanism-three-dimensional vibration data according to claim 1, characterized in that, In step 3), tool runout parameters are identified; The tool runout parameters are identified using an iterative optimization method. The specific steps are as follows: First, in the initial cutting stage, based on the milling force test data in the X and Y directions and the radial wear value collected by the force measuring instrument, the position angle of one revolution of the tool is selected as the sample point; Secondly, the initial values ​​of the tool radial runout and phase are set and substituted into the instantaneous cutting thickness model and cutting force model to calculate the theoretical cutting force value, which is then compared with the experimentally measured cutting force to form the sum of squared errors. Then, the iteration step size and runout parameter constraint range are set, and the parameter values ​​are continuously updated through iterative optimization to gradually reduce the difference between the theoretical and measured cutting forces. Finally, when the sum of squared errors reaches its minimum value, the corresponding radial runout and phase of the tool are the optimal identification results.

4. The method for real-time prediction of tool wear by fusing wear and vibration mechanism-three-dimensional vibration data according to claim 1, characterized in that, In step 4), the back face wear correction and the identified runout disturbance terms and cutting force are calculated in real time. The process involves introducing back face wear correction and the identified runout disturbance, specifically as follows: First, tool wear correction is introduced into the cutting force mechanism model constructed in step 2, taking into account the flank wear amount V. b The effect on effective cutting geometry increases Δh(V) b Correction of instantaneous cutting thickness: h (t)1 =h (t) +Δh(V b ) Wherein, Δh(V b () represents the wear correction amount; Then, the radial runout of the tool is introduced into the model, with the runout amount being Δr and the phase being φ, to correct the instantaneous cutting thickness h. (t)2 for: h (t)2 =h (t)1 +Δr·cos(wt+φ) Where w is the principal axis angular velocity; Finally, using the corrected instantaneous cutting thickness h (t)2 A cutting force prediction mechanism model that simultaneously considers tool wear and runout factors is constructed. The cutting force is calculated in real time. The specific operation is as follows: First, obtain the instantaneous cutting thickness h after considering tool wear and runout correction. (t)2 Then, the cutting forces are substituted into the milling force prediction mechanism model to calculate the three-dimensional cutting forces in real time during the machining process; Secondly, the forces acting on the discrete elements of the cutting edge are integrated using the infinitesimal element method to obtain the tangential force, radial force, and axial force components:

5. Among them, , , Indicates the tangential, radial, and axial cutting force components. , , Expressed as the cutting force coefficient; Finally, the force results of all cutting edges are superimposed to obtain the cutting force prediction data that takes into account the effects of tool wear and runout.

6. The method for real-time prediction of tool wear by fusing wear and vibration mechanism-three-dimensional vibration data according to claim 1, characterized in that, In step 5), a comprehensive feature vector for tool wear prediction is constructed; The collected three-dimensional vibration signals and the cutting force signals output by the mechanism model are preprocessed to extract time-domain, frequency-domain, and time-frequency-domain features respectively. Features highly correlated with tool wear are selected through correlation analysis. Finally, the vibration features and cutting force features are fused to form a comprehensive feature vector.

7. The method for real-time prediction of tool wear by fusing wear and vibration mechanism-three-dimensional vibration data according to claim 1, characterized in that, In step 6), the tool wear prediction model adopts a regression structure based on deep learning and a neural network with the ability to model time-series features, which is either a Long Short-Term Memory Network (LSTM) or a Convolutional Neural Network (CNN).

Images