Cutting path virtual simulation method based on digital twin
By using a digital twin-based virtual simulation method for cutting paths, the cutting path is decomposed into cutting segments, a neural network model is constructed, and parameters are iteratively adjusted. This solves the shortcomings of abnormal state identification and optimization in traditional cutting processes, and achieves efficient and stable machining results.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- WUXI HONGYING PRECISION TECH CO LTD
- Filing Date
- 2025-07-09
- Publication Date
- 2026-05-26
AI Technical Summary
Traditional cutting process monitoring methods struggle to identify and dynamically respond to abnormal states during the cutting process in real time, such as tool wear and cutting vibration, leading to low processing efficiency and unstable product quality. They also lack comprehensive evaluation of multiple signals and parameters, and traditional iterative optimization methods lack effective constraints.
A cutting path virtual simulation method based on digital twins is adopted. By collecting abnormal states, breaking down the cutting path into cutting segments, generating label data, constructing a neural network algorithm model, iteratively adjusting the cutting equipment parameters, and combining real-time feedback to optimize the CNC program, the method can achieve accurate identification and dynamic adjustment of abnormal states.
It enables accurate identification and dynamic response to various abnormal states, improves processing efficiency and quality stability, reduces energy consumption and extends tool life, and enhances the realism and reliability of digital twin models.
Smart Images
Figure CN120874534B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of path simulation technology, specifically a virtual simulation method for cutting paths based on digital twins. Background Technology
[0002] With the development of intelligent manufacturing, the manufacturing industry is paying more and more attention to digitalization and automation. Digital twins, as a cutting-edge technology, can simulate and optimize the manufacturing process in a virtual environment, thereby improving production efficiency and quality. The processing environment of modern manufacturing is becoming increasingly complex, involving the cutting of various materials and structures. Abnormal conditions such as tool wear and cutting vibration often occur during the processing, which can affect processing quality and production efficiency. In cutting, even small changes in equipment parameters can significantly affect the quality of the final product.
[0003] Traditional cutting process monitoring methods often struggle to identify and dynamically respond to various abnormal states that occur during the cutting process in real time, such as tool wear and cutting chatter. They lack dynamic feedback and flexibility, which can easily lead to low processing efficiency and unstable product quality. When dealing with the impact of abnormal states, they lack a comprehensive evaluation of multiple signals and parameters, making it impossible to fully quantify the impact of abnormal states on processing quality. Traditional iterative optimization methods lack effective constraints, and irrational parameter fluctuations can easily occur during the adjustment process. Summary of the Invention
[0004] (a) Technical problems to be solved
[0005] In view of the above-mentioned shortcomings of the existing technology, the present invention provides a cutting path virtual simulation method based on digital twin, which can effectively solve the problems of the existing technology.
[0006] (II) Technical Solution
[0007] To achieve the above objectives, the present invention provides the following technical solution:
[0008] This invention discloses a cutting path virtual simulation method based on digital twins, comprising the following steps:
[0009] Step 1: Collect or actively set several abnormal states that occur during the processing of the target cutting path, decompose the target cutting path into one or more continuous cutting segments according to the processing sequence, perform quantitative analysis on each cutting segment, generate corresponding label data, and predefine the benchmark cutting equipment operating parameters required to achieve abnormal cutting for each cutting segment based on the label data.
[0010] Step 2: Extract the impact features of each abnormal state defined in Step 1 on the cutting process. Based on the impact features, dynamically mark the cutting segments in the target cutting path affected by the abnormal state, and output the impact coefficient characterizing the severity of the abnormal state. Record the operating parameters of the cutting equipment corresponding to the marked abnormal state.
[0011] Step 3: Construct a simulation adjustment model based on a neural network algorithm. Input the label data of the marked cutting segment, the influence coefficient, and the corresponding operating parameters of the cutting equipment at the time of labeling into the simulation adjustment model. The simulation adjustment model iteratively outputs the operating parameters of the cutting equipment for adjusting a certain cutting segment according to the preset adjustment rules.
[0012] Step 4: After each parameter adjustment, re-evaluate the virtual machining process of the cutting segment and output the updated influence coefficient;
[0013] Step 5: Repeat the iterative adjustment process of Steps 3-4 until a preset stopping condition is met. Determine the operating parameters of the cutting equipment corresponding to the stopping condition as the optimal adjustment parameters for the abnormal marked cutting segment under the current abnormal state.
[0014] Step 6: Replace the reference cutting equipment operating parameters corresponding to the cutting segment with the optimal adjustment parameters obtained in Step 5, and execute Steps 4-6 on all marked cutting segments in the target cutting path to generate an optimized CNC program containing the corrected operating parameters;
[0015] Step 7: Execute the optimized CNC program on the machining equipment and record the feedback.
[0016] Furthermore, in step 1, when the target cutting path is decomposed into cutting segments, it is automatically divided based on the preset fixed path length in the CNC program, and the cutting segment boundary is automatically adjusted when the length difference between adjacent cutting segments exceeds a preset threshold.
[0017] Furthermore, the abnormal states in step 1 include: excessive tool wear, tool chipping, cutting chatter, internal defects in the workpiece material, loose fixture, and insufficient cooling; the tag data includes the cutting segment length, cutting depth, nominal feed rate, nominal spindle speed, and tool feed direction.
[0018] Furthermore, the influence characteristics of step 2 include: vibration energy changes in a specific frequency band, cutting force waveform distortion characteristics, abrupt change modes of acoustic emission signals, and abnormal temperature gradients.
[0019] Furthermore, the process of obtaining the abnormal state influence coefficient in step 2 is as follows:
[0020] Based on the vibration energy change characteristics corresponding to the abnormal state, the energy distribution offset within a specific frequency band is captured and converted into a standardized vibration deviation index.
[0021] Based on the characteristics of cutting force waveform distortion, the amplitude deviation and time-domain distribution characteristics of abnormal peaks and troughs in the waveform are analyzed to generate a cutting force distortion intensity factor.
[0022] For abrupt changes in acoustic emission signal patterns, the duration, peak energy, and frequency domain spread of sudden high-energy pulses are identified, and the acoustic emission abrupt change energy density is calculated.
[0023] For abnormal temperature gradients, extract the extreme values of temperature difference and spatial diffusion rate in the temperature field of the cutting area where the temperature rises or falls sharply, and form a temperature gradient offset coefficient.
[0024] The standardized vibration deviation index, cutting force distortion intensity factor, acoustic emission mutation energy density, and temperature gradient offset coefficient are normalized and mapped to a unified severity measurement space. Based on the weight ratio of each feature's influence on machining quality, the normalized feature values are weighted and fused to generate an initial influence coefficient that reflects the comprehensive influence of a single abnormal state.
[0025] The real-time operating parameters of the cutting equipment recorded when the abnormal state is associated with the abnormal state marker include spindle speed, feed rate and depth of cut. Based on the deviation between the real-time operating parameters and the reference parameters, the initial influence coefficient is corrected by working condition offset compensation, and finally the influence coefficient representing the actual severity of the current abnormal state is output.
[0026] Furthermore, the neural network algorithm in step 3 is a cascaded structure of a multi-layer convolutional neural network and a long short-term memory network. The multi-layer convolutional neural network is used to extract spatial features, the long short-term memory network is used to capture temporal correlations in the segment sequence, and a preset adjustment rule is jointly applied at the network output layer. The preset adjustment rule is used to constrain the search direction and adjustment step size of the neural network in the parameter space.
[0027] Furthermore, the preset adjustment rules in step 3 include: the spindle speed adjustment range does not exceed ±10%; the feed rate adjustment step size does not exceed ±15% of the nominal speed; the tool cutting depth adjustment does not exceed 0.1 mm per step; and only a single parameter change is allowed in the same iteration process.
[0028] Furthermore, the formula for calculating the influence coefficient is as follows:
[0029] ;
[0030] In the formula, This represents the overall impact coefficient of the abnormal state under the current iteration parameters. Representing the Each feature weight coefficient Representing the Original values of anomaly features Representing the first in historical data The minimum value of each feature. Representing the first in historical data The maximum value of each feature, Representing the Individual working condition compensation weight coefficient Representing the The deviation rate of each real-time operating parameter relative to the baseline parameter. and Both represent index variables of the summation term.
[0031] Furthermore, the stopping conditions for step 5 include: the updated influence coefficient reaching below a preset minimum threshold; when the influence coefficient decreases by less than a preset minimum rate of change for three consecutive iterations, the influence coefficient is determined to have converged and the iteration is stopped; and the preset maximum number of iterations is reached.
[0032] Furthermore, during the feedback recording process in step 7, the actual state data of the physical processing process is collected in real time to identify whether the expected abnormal state or new abnormality occurs. The workpiece quality, processing efficiency and equipment status in the actual state data are compared with the virtual simulation prediction results. If there is a deviation higher than the preset range, the actual abnormality marker, influence coefficient, operating parameters and processing results are re-extracted based on the actual state data and fed back to the simulation adjustment model to update the weight parameters of the neural network model or optimize the preset adjustment rules.
[0033] (III) Beneficial Effects
[0034] Compared with the known prior art, the technical solution provided by this invention has the following beneficial effects:
[0035] By analyzing the characteristics of various processing data, the impact of various abnormal states on the cutting process can be effectively quantified. Combined with real-time working condition data, dynamic influence coefficient correction is performed to ensure the accuracy of anomaly identification. It can reflect the actual processing status in real time, thereby achieving precise anomaly monitoring and intelligent adjustment, and enhancing the realism and reliability of the digital twin model.
[0036] By employing a multi-layer convolutional neural network and a long short-term memory network in series, spatial and temporal features can be extracted simultaneously, effectively capturing the sequence relationship and spatial changes of the cutting segments. The combination of deep learning models and preset adjustment rules makes parameter adjustment highly intelligent and adaptable, and can automatically optimize cutting parameters according to different abnormal states, thereby achieving the goals of improving processing efficiency, reducing energy consumption, and extending tool life.
[0037] By introducing iterative adjustments, automatic stopping conditions, and feedback mechanisms, the simulation adjustments are ensured to continuously approach the optimal solution and are verified in actual processing. By continuously re-evaluating the influence coefficients, adjusting parameters, and combining actual feedback data, deviations between virtual simulation and actual processing are effectively avoided, improving the overall stability and robustness of the process. Through its systematic and dynamic optimization characteristics, it provides a strong guarantee for high reliability and high efficiency in industrial applications. Attached Figure Description
[0038] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.
[0039] Figure 1 This is a schematic diagram of the process of the present invention. Detailed Implementation
[0040] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.
[0041] The present invention will be further described below with reference to embodiments.
[0042] Example 1
[0043] The cutting path virtual simulation method based on digital twins in this embodiment, such as Figure 1 As shown, it includes the following steps:
[0044] Step 1: Collect or actively set several abnormal states that may occur during the processing of the target cutting path. Decompose the target cutting path into one or more continuous cutting segments according to the processing sequence. Quantify and analyze each cutting segment to generate corresponding label data. Based on the label data, predefine the benchmark cutting equipment operating parameters required to achieve abnormal-free cutting for each cutting segment. When decomposing the target cutting path into cutting segments, automatically divide it based on the preset fixed path length in the CNC program. When the length difference between adjacent cutting segments exceeds a preset threshold, automatically adjust the cutting segment boundaries. Ensure that the cutting segment division is both detailed and uniform, facilitating subsequent analysis and traceability. Predefine the ideal operating parameters for abnormal-free states and construct a control group to provide clear targets for subsequent adjustments.
[0045] Abnormal conditions include: excessive tool wear, tool chipping, cutting chatter, internal defects in workpiece material, loose fixture, and insufficient cooling; the tag data includes cutting segment length, cutting depth, nominal feed rate, nominal spindle speed, and tool feed direction.
[0046] Step 2: Extract the impact features of each abnormal state defined in Step 1 on the cutting process. Based on the impact features, dynamically mark the cutting segments in the target cutting path affected by the abnormal state and output the impact coefficient characterizing the severity of the abnormal state. Record the operating parameters of the cutting equipment corresponding to the marked abnormal state. The impact features include: vibration energy changes in specific frequency bands, cutting force waveform distortion characteristics, abrupt change mode of acoustic emission signal, and abnormal temperature gradient. Improve the sensitivity and coverage of abnormal state detection, output the impact coefficient, intuitively reflect the degree of impact of different abnormalities on cutting quality and equipment status, and archive the operating parameters at the time of marking to provide real working condition samples for subsequent model training and simulation adjustment.
[0047] Step 3: Construct a simulation adjustment model based on a neural network algorithm. Input the labeled data of the marked cutting segments, the influence coefficients, and the corresponding operating parameters of the cutting equipment at the time of labeling into the simulation adjustment model. The simulation adjustment model iteratively outputs the operating parameters of the cutting equipment for adjusting a certain cutting segment according to the preset adjustment rules. The neural network algorithm is a cascaded structure of a multi-layer convolutional neural network and a long short-term memory network. The multi-layer convolutional neural network is used to extract spatial features, and the long short-term memory network is used to capture the temporal correlation in the cutting segment sequence. The preset adjustment rules are combined in the network output layer to constrain the search direction and adjustment step size of the neural network in the parameter space. The multi-layer convolutional neural network extracts spatial features, and the LSTM captures the temporal correlation in the cutting segment sequence, making the model more comprehensive. The preset adjustment rules constrain the network search to ensure that the output parameters are within a safe and feasible range. Through multiple rounds of forward iteration, the optimal parameter configuration is quickly approximated, reducing the reliance on manual experience for parameter tuning.
[0048] The preset adjustment rules include: the spindle speed adjustment range shall not exceed ±10%; the feed rate adjustment step size shall not exceed ±15% of the nominal speed; the tool cutting depth adjustment shall not exceed 0.1 mm per step; and only a single parameter change is allowed in the same iteration process.
[0049] Step 4: After each parameter adjustment, re-evaluate the virtual processing of the cutting segment and output the updated influence coefficient. The evaluation is carried out in the digital twin environment immediately after each parameter adjustment, so that errors and deviations can be captured and corrected in a timely manner. The updated influence coefficient is continuously output to provide the latest feedback for subsequent iterations.
[0050] Step 5: Repeat the iterative adjustment process of Steps 3-4 until a preset stopping condition is met. The operating parameters of the cutting equipment corresponding to the stopping condition are determined as the optimal adjustment parameters for the abnormal marked cutting segment under the current abnormal state. The stopping conditions include: the updated influence coefficient reaches below the preset minimum threshold; when the decrease in influence coefficient for three consecutive iterations is less than the preset minimum rate of change, the influence coefficient is determined to have converged and the iteration is stopped; the preset maximum number of iterations is reached; and the process is terminated in place when there is no significant improvement in continuous iterations, to avoid invalid calculations and excessive parameter adjustment.
[0051] Step 6: Replace the reference cutting equipment operating parameters corresponding to the cutting segment with the optimal adjustment parameters obtained in Step 5. Perform Steps 4-6 on all marked cutting segments in the target cutting path to generate an optimized CNC program containing the corrected operating parameters. Replace the reference with the optimal adjustment parameters of all marked segments in batches to automatically generate an optimized NC program, covering all abnormal segments at once to ensure the continuity and stability of subsequent machining.
[0052] Step 7: Execute the optimized CNC program on the machining equipment and record the feedback. During the feedback recording process, collect the actual state data of the physical machining process in real time, identify whether the expected abnormal state or new abnormality occurs, and compare the workpiece quality, machining efficiency and equipment status in the actual state data with the virtual simulation prediction results. If there is a deviation greater than the preset range, then based on the actual state data, re-extract the actual abnormality marker, influence coefficient, running parameters and machining results, and feed them back to the simulation adjustment model to update the weight parameters of the neural network model or optimize the preset adjustment rules, thereby improving the prediction accuracy and optimization effect of subsequent simulations.
[0053] Compared with existing technologies, this method automatically decomposes the entire cutting path into quantified and labeled cutting segments, dynamically marks various typical anomalies and extracts their impact features, and then combines a multi-layer convolutional neural network and a long short-term memory network in series with preset adjustment rules to iteratively simulate and adjust the abnormal cutting segments until the impact coefficients converge or reach the optimal level. At the same time, closed-loop feedback is performed in actual machining to correct the model and rules online, thereby achieving accurate prediction and parameter optimization of cutting anomalies. On the one hand, it avoids the blindness and tediousness of traditional experience-based parameter tuning, and on the other hand, it significantly improves machining quality, efficiency and the reliability of digital twin simulation.
[0054] Example 2
[0055] At other levels, this embodiment also provides another optimization mechanism based on embodiment 1, specifically a process for obtaining the influence coefficient of abnormal states as follows:
[0056] Based on the vibration energy change characteristics corresponding to the abnormal state, the energy distribution offset within a specific frequency band is captured and converted into a standardized vibration deviation index.
[0057] Based on the characteristics of cutting force waveform distortion, the amplitude deviation and time-domain distribution characteristics of abnormal peaks and troughs in the waveform are analyzed to generate a cutting force distortion intensity factor.
[0058] For abrupt changes in acoustic emission signal patterns, the duration, peak energy, and frequency domain spread of sudden high-energy pulses are identified, and the acoustic emission abrupt change energy density is calculated.
[0059] For abnormal temperature gradients, extract the extreme values of temperature difference and spatial diffusion rate in the temperature field of the cutting area where the temperature rises or falls sharply, and form a temperature gradient offset coefficient.
[0060] The standardized vibration deviation index, cutting force distortion intensity factor, acoustic emission mutation energy density, and temperature gradient offset coefficient are normalized and mapped to a unified severity measurement space. Based on the weight ratio of each feature's influence on machining quality, the normalized feature values are weighted and fused to generate an initial influence coefficient that reflects the comprehensive influence of a single abnormal state.
[0061] The real-time operating parameters of the cutting equipment recorded when the abnormal state is associated with the abnormal state marker include spindle speed, feed rate and depth of cut. Based on the deviation between the real-time operating parameters and the reference parameters, the initial influence coefficient is corrected by working condition offset compensation, and finally the influence coefficient representing the actual severity of the current abnormal state is output.
[0062] Compared with existing technologies, this embodiment targets multiple sources of features such as vibration energy deviation, cutting force distortion, acoustic emission abrupt energy density, and temperature gradient deviation. It first extracts the standardized vibration deviation index, cutting force distortion intensity factor, acoustic emission abrupt energy density, and temperature gradient deviation coefficient, respectively. Then, it normalizes and maps these parameters to a unified severity metric space and weights them according to their influence weights. Finally, it combines the deviation between real-time operating parameters and benchmark parameters to perform working condition compensation correction. This achieves a high-precision quantitative assessment of the comprehensive impact of a single abnormal state, significantly improving the accuracy of abnormal severity determination and the reliability of parameter optimization adjustment compared with existing technologies.
[0063] Example 3
[0064] This embodiment provides a formula for calculating the influence coefficient, specifically:
[0065] ;
[0066] In the formula, This represents the overall impact coefficient of the abnormal state under the current iteration parameters. Representing the Each feature weight coefficient Representing the Original values of anomaly features Representing the first in historical data The minimum value of each feature. Representing the first in historical data The maximum value of each feature, Representing the Individual working condition compensation weight coefficient Representing the The deviation rate of each real-time operating parameter relative to the baseline parameter. and All represent index variables of the summation term, including: spindle speed deviation rate, feed rate deviation rate, and depth of cut deviation rate. Through a dual mechanism of multi-source feature fusion and dynamic working condition compensation, the intensity of physical signal anomalies is quantified, dimensional differences are eliminated by normalization, and weighted superposition reflects the comprehensive anomaly level. This enables the influence coefficient to dynamically respond to the actual working conditions of the equipment, avoids the disconnect between virtual simulation and physical execution, and improves the online correction accuracy of the model.
[0067] In summary, this invention achieves accurate identification and evaluation of abnormal states such as tool wear and vibration chatter by automatically segmenting and quantifying cutting segments based on CNC programs, and by extracting various working condition features and normalizing influence coefficients. On this basis, a simulation adjustment model is constructed by connecting a multi-layer convolutional neural network and a long short-term memory network, and the spindle speed, feed rate and cutting depth are iteratively optimized by using preset adjustment rules, which quickly converges to the optimal operating parameters.
[0068] The final result is an optimized CNC program containing correction parameters, and the model and rules are continuously updated through a real-time machining feedback loop, which effectively improves machining quality, efficiency and stability.
[0069] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions will not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A cutting path virtual simulation method based on digital twins, characterized in that, Includes the following steps: Step 1: Collect or actively set several abnormal states that occur during the processing of the target cutting path, decompose the target cutting path into one or more continuous cutting segments according to the processing sequence, perform quantitative analysis on each cutting segment, generate corresponding label data, and predefine the benchmark cutting equipment operating parameters required to achieve abnormal cutting for each cutting segment based on the label data. Step 2: Extract the impact features of each abnormal state defined in Step 1 on the cutting process. Based on the impact features, dynamically mark the cutting segments in the target cutting path affected by the abnormal state, and output the impact coefficient characterizing the severity of the abnormal state. Record the operating parameters of the cutting equipment corresponding to the marked abnormal state. Step 3: Construct a simulation adjustment model based on a neural network algorithm. Input the label data of the marked cutting segment, the influence coefficient, and the corresponding operating parameters of the cutting equipment at the time of labeling into the simulation adjustment model. The simulation adjustment model iteratively outputs the operating parameters of the cutting equipment for adjusting a certain cutting segment according to the preset adjustment rules. Step 4: After each parameter adjustment, re-evaluate the virtual machining process of the cutting segment and output the updated influence coefficient; Step 5: Repeat the iterative adjustment process of Steps 3-4 until a preset stopping condition is met. Determine the operating parameters of the cutting equipment corresponding to the time the stopping condition is reached as the optimal adjustment parameters of the cutting segment affected by the abnormal state under the current abnormal state. Step 6: Replace the reference cutting equipment operating parameters corresponding to the cutting segment with the optimal adjustment parameters obtained in Step 5, and execute Steps 4-6 on all marked cutting segments in the target cutting path to generate an optimized CNC program containing the corrected operating parameters; Step 7: Execute the optimized CNC program on the machining equipment and record the feedback; The process of obtaining the abnormal state influence coefficient in step 2 is as follows: Based on the vibration energy change characteristics corresponding to the abnormal state, the energy distribution offset within a specific frequency band is captured and converted into a standardized vibration deviation index. Based on the characteristics of cutting force waveform distortion, the amplitude deviation and time-domain distribution characteristics of abnormal peaks and troughs in the waveform are analyzed to generate a cutting force distortion intensity factor. For abrupt changes in acoustic emission signal patterns, the duration, peak energy, and frequency domain spread of sudden high-energy pulses are identified, and the acoustic emission abrupt change energy density is calculated. For abnormal temperature gradients, extract the extreme values of temperature difference and spatial diffusion rate in the temperature field of the cutting area where the temperature rises or falls sharply, and form a temperature gradient offset coefficient. The standardized vibration deviation index, cutting force distortion intensity factor, acoustic emission mutation energy density, and temperature gradient offset coefficient are normalized and mapped to a unified severity measurement space. Based on the weight ratio of each feature's influence on machining quality, the normalized feature values are weighted and fused to generate an initial influence coefficient that reflects the comprehensive influence of a single abnormal state. The real-time operating parameters of the cutting equipment recorded when the abnormal state is associated with the abnormal state marker include spindle speed, feed rate and depth of cut. Based on the deviation between the real-time operating parameters and the reference parameters, the initial influence coefficient is corrected by working condition offset compensation, and finally the influence coefficient representing the actual severity of the current abnormal state is output.
2. The virtual simulation method for cutting paths based on digital twins according to claim 1, characterized in that, In step 1, when the target cutting path is decomposed into cutting segments, it is automatically divided based on the preset fixed path length in the CNC program, and the cutting segment boundary is automatically adjusted when the length difference between adjacent cutting segments exceeds a preset threshold.
3. The virtual simulation method for cutting paths based on digital twins according to claim 1, characterized in that, The abnormal states in step 1 include: excessive tool wear, tool chipping, cutting chatter, internal defects in the workpiece material, loose fixture, and insufficient cooling; the tag data includes the cutting segment length, cutting depth, nominal feed rate, nominal spindle speed, and tool feed direction.
4. The virtual simulation method for cutting paths based on digital twins according to claim 1, characterized in that, The influence characteristics of step 2 include: vibration energy changes in a specific frequency band, cutting force waveform distortion characteristics, abrupt change modes of acoustic emission signals, and abnormal temperature gradients.
5. The virtual simulation method for cutting paths based on digital twins according to claim 1, characterized in that, The neural network algorithm in step 3 is a cascaded structure of a multi-layer convolutional neural network and a long short-term memory network. The multi-layer convolutional neural network is used to extract spatial features, and the long short-term memory network is used to capture temporal correlations in the segment sequence. A preset adjustment rule is jointly set at the network output layer. The preset adjustment rule is used to constrain the search direction and adjustment step size of the neural network in the parameter space.
6. The virtual simulation method for cutting paths based on digital twins according to claim 1, characterized in that, The preset adjustment rules in step 3 include: the spindle speed adjustment range shall not exceed ±10%; the feed rate adjustment step size shall not exceed ±15% of the nominal speed; the tool cutting depth adjustment shall not exceed 0.1 mm per step; and only a single parameter change is allowed in the same iteration process.
7. The virtual simulation method for cutting paths based on digital twins according to claim 1, characterized in that, The formula for calculating the influence coefficient is as follows: ； In the formula, This represents the overall impact coefficient of the abnormal state under the current iteration parameters. Representing the Each feature weight coefficient Representing the Original values of anomaly features Representing the first in historical data The minimum value of each feature. Representing the first in historical data The maximum value of each feature, Representing the Individual working condition compensation weight coefficient Representing the The deviation rate of each real-time operating parameter relative to the baseline parameter. and Both represent index variables of the summation term.
8. The virtual simulation method for cutting paths based on digital twins according to claim 1, characterized in that, The stopping conditions for step 5 include: The updated impact coefficient has fallen below the preset minimum threshold; When the decrease in the influence coefficient is less than the preset minimum rate of change after three consecutive iterations, it is determined that the influence coefficient has converged and the iteration is stopped. The preset maximum number of iterations has been reached.
9. The virtual simulation method for cutting paths based on digital twins according to claim 1, characterized in that, In step 7, during the feedback recording process, the actual state data of the physical processing process is collected in real time to identify whether the expected abnormal state or new abnormality occurs. The workpiece quality, processing efficiency and equipment status in the actual state data are compared with the virtual simulation prediction results. If there is a deviation greater than the preset range, the actual abnormality marker, influence coefficient, operating parameters and processing results are re-extracted based on the actual state data and fed back to the simulation adjustment model to update the weight parameters of the neural network model or optimize the preset adjustment rules.