Circumferential contour prediction method based on cutting force and deformation field space-time correlation

By constructing a lightweight stiffness matrix and an elastic mechanism model, and combining it with measured signals for state correction, the problem of efficient and accurate prediction of dynamic deformation field in the machining of the circumferential contour of annular thin-walled structural parts was solved. This enabled real-time deformation field prediction at the workstation level, improving manufacturing quality and production efficiency.

CN121683069APending Publication Date: 2026-03-17SHANGHAI JIAOTONG UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202511758547.2
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2025-11-26
Publication Date
2026-03-17

AI Technical Summary

Technical Problem

Existing technologies struggle to achieve efficient and accurate prediction of dynamic deformation fields in the machining of the circumferential contour of annular thin-walled structures, especially under high temperature, high pressure, and high speed conditions. Existing methods suffer from high computational complexity and slow response speed, making it difficult to meet the needs of industrial sites.

Method used

A prediction method based on the spatiotemporal correlation of cutting force and deformation field is adopted. By constructing a lightweight structural stiffness matrix, an elastic mechanism model and a hybrid neural network, and combining it with measured process signals for state correction, efficient and accurate prediction of the deformation field is achieved.

Benefits of technology

It significantly improves the accuracy and response speed of online prediction, enabling real-time prediction of circumferential deformation fields at the workstation level, thereby improving manufacturing quality and production efficiency. It is suitable for annular thin-walled structural components in high-end manufacturing fields.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN121683069A_ABST
    Figure CN121683069A_ABST
Patent Text Reader

Abstract

The invention discloses a circumferential contour prediction method based on cutting force and deformation field space-time correlation, and relates to the field of aerospace, and the method comprises the following steps: 1, carrying out the spatial discretization of a circumferential contour region of a target workpiece; step 2, establishing a lightweight structure stiffness matrix; 3, constructing an elastic mechanism model and solving delta-F mapping; 4, introducing the time sequence correlation of the cutting load into the elastic mechanism model; step 5, constructing a Boolean relationship between the tool path and the workpiece envelope; step 6, fusing actual measurement process signals, and performing state correction; step 7, iterating the elastic model again by using the corrected load; and step 8, reconstructing full-field deformation based on spatial correlation. According to the method, online prediction of the station-level circumferential deformation field is achieved, the method can be used in the multi-procedure machining process, deformation in-situ compensation of the procedure is achieved, a clamping positioning strategy of the next procedure is output, and the manufacturing quality and the production efficiency are improved.
Need to check novelty before this filing date? Find Prior Art

Description

TECHNICAL FIELD

[0001] The present application relates to the field of aerospace, and in particular to a circumferential profile prediction method based on cutting force and deformation field space-time correlation. BACKGROUND

[0002] Annular thin-walled structures are widely used in aerospace equipment, typical components including gas turbine casings, integrated blisks, bearings, cabin bodies and fuselage frames, etc. Such parts are usually in complex operating conditions of high temperature, high pressure, high speed and high sealing, which puts high requirements on manufacturing precision, stability and geometric consistency. For example, gas turbine casings are often made of high-strength, low-rigidity materials such as titanium alloy, stainless steel and high-temperature alloy, which are difficult to machine. In the cutting process, geometric errors are easily caused by elastic deformation, which becomes the main source of part rejection and production efficiency bottleneck.

[0003] In the manufacturing process, due to the coupling of factors such as significant cutting load, uneven material removal, and strong time-varying workpiece stiffness, annular structures are prone to produce non-uniform and nonlinear deformation fields in machining. Especially in the circumferential direction, the complex force-thermal-structure coupling effect significantly aggravates the difficulty of deformation prediction. Existing deformation prediction techniques can be roughly divided into three categories: finite element method, mechanism model and data-driven model. The finite element method has strong physical modeling capability and can consider multiple factors such as material damage, residual stress evolution and structural stiffness change, but the model construction and solution process is computationally expensive, making it difficult to meet the requirements of prediction efficiency and response time in industrial field. The mechanism model can construct the corresponding deformation mapping function through elastic stiffness modeling, load modeling and residual stress modeling, and has good interpretability and expandability under certain conditions, but it is difficult to fully represent the nonlinear behavior under complex working conditions. Data-driven models rely on process signals (such as cutting force, power, vibration, etc.) to construct end-to-end prediction mapping, and have high accuracy, but face problems such as strong sample dependence and weak generalization ability.

[0004] In summary, for the circumferential profile machining of annular structures, the existing research faces the following key technical difficulties: 1. The machining objects are mostly high-value small-batch parts, making it difficult to obtain large-scale samples to meet the training needs of data-driven models; 2. The circumferential feature is extended from two-dimensional plane to three-dimensional surface, the deformation mode is more complex, the stiffness matrix dimension is increased, resulting in a sharp increase in the complexity of deformation field prediction calculation; 3. The mechanism model is not sufficient to represent complex working conditions, while the monitoring signals are rich in information, but the fusion mechanism between physical modeling is still immature; 4. The machining process has significant nonlinear time-varying and spatial correlation, and existing methods are difficult to quickly respond to prediction needs in online state.

[0005] Therefore, the person skilled in the art is committed to developing a deformation prediction method for ring-shaped thin-walled parts with precision and efficiency, realizing efficient and accurate prediction of dynamic deformation field. SUMMARY

[0006] In view of the above defects of the prior art, the technical problem to be solved by the present application is how to realize efficient and accurate prediction of dynamic deformation field for ring-shaped thin-walled parts.

[0007] To achieve the above-mentioned purpose, the present application provides a circumferential contour prediction method based on cutting force and deformation field space-time correlation, characterized in that the method comprises the following steps: Step 1, spatial discretization of the circumferential contour region of the target workpiece; Step 2, establishing a lightweight structure stiffness matrix; Step 3, constructing an elastic mechanism model and solving the δ-F mapping, wherein represents the resultant load on the node, represents the elastic deformation variable at the node; Step 4, introducing the time sequence correlation of the cutting load in the elastic mechanism model; Step 5, constructing the Boolean relationship between the tool path and the workpiece envelope, and extracting the nominal cutting depth as the initial value for solving the elastic mechanism model in step 3; Step 6, fusing the measured process signals and correcting the state; Step 7, using the corrected load to re-iterate the elastic model to obtain the corrected deformation value; Step 8, reconstructing the full-field deformation based on spatial correlation.

[0008] Further, the step 1 further comprises: adopting a meshing method to perform surface subdivision on the circumferential contour region of the target workpiece, constructing a discrete node set in a three-dimensional coordinate system, each node representing a local deformation point to be predicted, and the discrete strategy adopts a "layer-angle" discrete strategy in a polar coordinate system to perform spatial subdivision on the target circumferential deformation region.

[0009] Further, the step 2 further comprises: Step 2.1, based on the geometric envelope relationship between the cutting tool and the target workpiece, spatially partitioning the target workpiece to identify the cutting layer and the non-cutting layer; Step 2.2, using finite element analysis software to model the structure of the model after spatial partitioning, introducing a substructure method to combine the non-cutting layer region into an equivalent super-element, and realizing the stiffness compression representation of the non-processing region.

[0010] Further, the step 2 further comprises: Step 2.3, establishing the overall structure stiffness matrix of the target workpiece in the unprocessed state and the local stiffness matrix collection of all elements in the cutting area, determining the element set of material removal by identifying the intersection of the envelope surface of the cutting tool and the target workpiece at each discrete time step, and using the stiffness matrix reduction technique to eliminate the local stiffness matrix collection of all elements in the cutting area from the overall structure stiffness matrix to obtain the structure stiffness matrix at each discrete time.

[0011] Further, the step 3 further comprises: based on the relationship of elasticity mechanics = · , establishing the coupling model between the resultant load on the node and the elastic deformation variable at the node, wherein is the structure stiffness matrix at each discrete time.

[0012] Further, the step 3 further comprises constructing a numerical model based on the iterative approximation among “force-deformation-cutting depth”, and the iterative process of the first node is as follows: Step 3.1, initialization, setting the initial cutting depth and the initial elastic deformation variable, and calculating the initial cutting force from the cutting force empirical model; Step 3.2, linear solution, calculating the node deformation by using the current load and the structure stiffness matrix at each discrete time; Step 3.3, depth correction, updating the cutting depth according to the deformation result, and recalculating the cutting load accordingly; Step 3.4, convergence judgment, if the difference between the deformation of the first node and the deformation of the first node is less than the given iteration accuracy, it is considered that the balance is reached, and the iteration is ended, otherwise, it returns to step 3.2 for continuous iteration.

[0013] Further, the step 4 further comprises modeling the duration cutting force data by using the hybrid neural network in the continuous machining process, and the output result is used as the iterative initial value of the initial elastic deformation variable in step 3.1.

[0014] Further, in the step 6, the measured process signal is the measured value of the cutting load.

[0015] Further, the step 6 further comprises: Step 6.1, constructing the system state equation for the cutting deformation and the cutting force balance process at each sampling point of the global deformation; Step 6.2, constructing the observation equation: , wherein, is the cutting force observation vector at the sampling point k. For the measurement matrix, It is a white noise vector. This is the calculated cutting force state vector at sampling point k.

[0016] Furthermore, in step 8, the full-field deformation adopts a sparse sampling scheme, with the input being the spatial location and deformation value of the sparse sampling points, and the output being the deformation estimate of the entire surface node.

[0017] The beneficial technical effects of the present invention are as follows: 1. Circular contour cutting deformation is strongly influenced by multiple coupled factors, and existing models have limited prediction accuracy, making it difficult to achieve online deformation field prediction at the workstation. This invention constructs a hybrid-driven prediction model based on the temporal correlation of cutting loads and the spatial correlation of the deformation field. It models the mapping relationship between δ and FK during the machining process based on the elastic deformation mechanism, analyzes the temporal correlation of cutting loads to accelerate the convergence of the mechanism model, and integrates the temporal information of measured cutting loads to improve model accuracy. This significantly improves online prediction accuracy and response speed, achieving real-time prediction of the circumferential deformation field at the workstation level. It can be used in multi-process machining to achieve in-situ deformation compensation for the current process and output clamping and positioning strategies for the next process, improving manufacturing quality and production efficiency.

[0018] 2. The stiffness matrix of ring-shaped structural components is time-varying, nonlinear, and high-dimensional, resulting in significant time consumption for modeling and solving, making it difficult to apply to real-time computing scenarios. This invention proposes a lightweight representation and modeling method for time-varying stiffness matrices. By simplifying the stiffness matrix, decomposing the domain, and modeling locally, its dimensionality and update complexity are significantly reduced, achieving efficient solutions under time-varying modeling. This reduces the computational complexity of the model and enables sub-second-level solutions on ordinary industrial computing platforms, making it suitable for online applications.

[0019] 3. Cutting deformation exhibits a significant time-cumulative effect. Traditional models suffer from low iteration efficiency and slow convergence speed. This invention establishes a spatiotemporal correlation analysis model for cutting loads, enabling rapid δ-F iteration. By analyzing the temporal correlation of cutting load signals, key dynamic features are extracted, providing optimal initial values ​​for iterative solution of the deformation model. This shortens the model convergence time, improves prediction efficiency, and meets the real-time processing requirements in dynamic scenarios.

[0020] 4. The limited number of monitoring points during processing cannot cover all deformation information, affecting prediction accuracy. This invention proposes a sparse sampling strategy and reconstruction algorithm for the deformation field. Based on compressed sensing and interpolation reconstruction methods, it reconstructs the overall deformation field from a limited number of sampling points, improving data utilization. Under the premise of significantly reducing the number of measurements and measurement costs, it achieves high-quality global deformation prediction, has good engineering adaptability, and is conducive to engineering application.

[0021] 5. Data-driven models have poor generalization ability and weak physical interpretability, making them difficult to reuse across different operating conditions. This invention adopts a hybrid modeling strategy of "mechanism model as the main approach and data-driven approach as the auxiliary approach." The mechanism model provides structural constraints and physical meanings of parameters, while the data-driven model fills in the gaps in unmodeled mechanism information, improving the model's completeness and robustness. The predictive model has stronger versatility and transferability, and can adapt to different part materials and part topologies.

[0022] The following will further explain the concept, specific structure, and technical effects of the present invention in conjunction with the accompanying drawings, so as to fully understand the purpose, features, and effects of the present invention. Attached Figure Description

[0023] Figure 1 This is a flowchart of a preferred embodiment of the present invention for predicting a circular profile based on the spatiotemporal correlation of cutting force and deformation field. Figure 2 This is a preferred embodiment of the tool and workpiece contact iteration process of the present invention; Figure 3 This is an iterative convergence equilibrium process for the deformation mechanism model of a preferred embodiment of the present invention; Figure 4 This is a flowchart of cutting load timing correlation extraction according to a preferred embodiment of the present invention; Figure 5 This is a preferred embodiment of the cutting deformation field spatial correlation extraction process of the present invention; Figure 6 This is a flowchart illustrating the operation of a prediction model deployed in the edge system of a machine tool equipment according to a preferred embodiment of the present invention. Figure 7 This is a preferred embodiment of the workpiece clamping boundary condition of the present invention; Among them, 1-workpiece, 2-turning tool, 3-Y-direction limiting support, 4-X-direction limiting support, 5-main support surface of fixture, 6-clamping block, 7-tensioning device, 202-base surface, 204-main cutting edge, 205-secondary cutting edge, 206-center point of cutting edge arc, 301-model input, 302-initial setting of iterative cycle, 303-calculate cutting force, 304-calculate new deformation, 305-calculate updated depth of cut, 306-determine whether the cycle upper limit has been reached, 307-determine whether convergence has been achieved, 308-model output. Detailed Implementation

[0024] The following description, with reference to the accompanying drawings, illustrates several preferred embodiments of the present invention to make its technical content clearer and easier to understand. The present invention can be embodied in many different forms, and the scope of protection of the present invention is not limited to the embodiments mentioned herein.

[0025] In the accompanying drawings, components with the same structure are indicated by the same numerical designation, and components with similar structures or functions are indicated by similar numerical designations. The dimensions and thicknesses of each component shown in the drawings are arbitrary, and the present invention does not limit the dimensions and thicknesses of each component. To make the illustrations clearer, the thickness of some components has been appropriately exaggerated in the drawings.

[0026] This invention provides a method for online prediction of the deformation field during the machining of the circumferential contour of a thin-walled annular component. It integrates an elasticity model with cutting process data, achieving real-time deformation prediction during machining through efficient modeling and data fusion. This method is applicable to the machining of the inner and outer circular contours of thin-walled circular parts made of high-strength, low-modulus materials such as titanium alloys and high-temperature alloys. The workpiece clamping boundary conditions are as follows: Figure 7 As shown. Figure 1 As shown, the present invention specifically includes the following steps: Step 1: Spatial discretization of the target deformation field For the circumferential contour region of the target workpiece, a meshing method is used to perform surface subdivision, constructing a discrete node set {( , Each node represents a local deformation point to be predicted. This discrete structure provides the basic framework for subsequent deformation modeling in terms of spatial solution. To meet the prediction accuracy requirements, a layer-corner discretization strategy in polar coordinates is used to spatially partition the target circumferential deformation region. This discretization strategy is based on the spatial distribution characteristics of deformation during processing and matches the minimum resolution of the coordinate measuring system to ensure that the input and output accuracy of the prediction model meets the actual detection accuracy requirements. In addition, combined with cutting process parameters, a spatiotemporal correlation mapping of the processing process is constructed on the discrete nodes to achieve integrated modeling of spatial discretization and time dimension.

[0027] Step 2: Establish the stiffness matrix of the lightweight structure

[0028] like Figure 2 As shown, a lightweight stiffness matrix construction method combining regional meshing and substructure method is used to realize dynamic modeling of workpiece stiffness during machining. First, based on the geometric envelope relationship between the tool and the workpiece, the target workpiece is spatially partitioned to identify the cutting layer and the non-cutting layer. In the meshing stage, a "fine + coarse" hybrid strategy is adopted: fine mesh modeling is used for the machining area (cutting layer) to improve the local prediction accuracy; coarse mesh modeling is used for the non-machining area (non-cutting layer) to reduce the overall computational load. Subsequently, the finite element analysis software is used to perform structural modeling on the partitioned model, and the substructure method is introduced to merge the non-cutting layer area into an equivalent super element to realize the stiffness compression representation of the non-machining area. In this process, two types of core data are output: (1) the overall structural stiffness matrix of the workpiece in the unmachined state. The matrix contains key physical properties such as material parameters and clamping boundary conditions; (2) the set of local stiffness matrices of all elements in the cutting region { },in Represents the local stiffness matrix of the i-th cutting element, with the entire set covering all effective elements in the fine mesh region. Each All data are extracted using a finite element solver, containing the material properties, geometric information, and boundary coupling relationships of the corresponding elements, and can be stitched into the overall structural stiffness matrix in the global coordinate system. This matrix set serves not only as the input data source for deformation calculations but also for stiffness update operations during the material removal process.

[0029] During the simulated machining process, the set of cells for material removal {e_remove} is determined by identifying the intersection of the tool and workpiece envelope at each discrete time step. }. Combined with { }, using stiffness matrix reduction technology, the corresponding From the original structural stiffness matrix The structural stiffness matrix at each moment is obtained by removing out faulty components and then dynamically updating it. This enables the modeling of stiffness evolution as processing progresses.

[0030] Step 3: Construct an elasticity mechanism model and solve the δ-F mapping.

[0031] Obtain the structural stiffness matrix at each discrete time point. Then, based on the relationship of elasticity: = · A coupling model between the applied load and the elastic deformation of the workpiece during the cutting process is established, in which... This represents the resultant force load at the node. This represents the elastic deformation at the corresponding node.

[0032] When predicting deformation behavior during machining, the model needs to consider the following coupling mechanism: local deformation of the workpiece will cause changes in the actual cutting depth, which in turn affects the magnitude of the cutting load; and the load will act on the workpiece through the structural stiffness matrix, causing new deformation.

[0033] like Figure 3 As shown, to solve this coupled nonlinear equilibrium relationship, a numerical model based on iterative approximation between "force-deformation-depth of cut" is constructed. The iteration process for each node is as follows: 1. Initialization: Set the initial cutting depth and initial elastic deformation The initial cutting force is calculated using an empirical model of cutting force. ; 2. Linear solution: using the current load and structural stiffness matrix Calculate node deformation ; 3. Depth Correction: Update the cutting depth based on the deformation results. = And recalculate the cutting load accordingly. = ( ); 4. Convergence Criterion: If If the condition is met, then the iteration ends; otherwise, return to step 2 and continue iterating.

[0034] This iterative framework enables dynamic coupling between workpiece local deformation, load, and cutting depth, allowing the prediction model to accurately capture load changes caused by elastic deformation, thereby improving the accuracy and reliability of deformation prediction.

[0035] Step 4: Introduce the temporal correlation of cutting loads

[0036] like Figure 4 As shown, the initialization input in step 3 Based on the fitting in this step, it is considered that during continuous machining, the cutting force at the previous moment has a significant impact on the mechanical state at the next moment. A CNN-LSTM hybrid neural network is used to analyze the historical cutting force data. The model is structured into four layers: an input layer, a CNN layer, an LSTM layer, and an output layer. The input layer preprocesses the raw data, the CNN layer extracts temporal features, the LSTM layer learns the features extracted by the CNN layer, and the output layer restores the LSTM predictions to the same data format as the model input. The model output is... These values ​​serve as initial values ​​for iterative calculations of the mechanistic model and for solving the elasticity model, improving model convergence speed and enhancing stability. Model efficiency requirements outweigh accuracy requirements, and the model structure needs to be shallow and have a fast response.

[0037] Step 5: Construct the Boolean relationship between the tool path and the workpiece envelope, and extract the nominal depth of cut.

[0038] Step 3 Initialization Input This step involves fitting the toolpath generated from the G-code and performing Boolean operations on the workpiece's 3D model to extract the nominal depth of cut. , as the initial values ​​for solving the elasticity model.

[0039] Step 6: Integrate the measured process signals and perform state correction.

[0040] By introducing the real-time monitored cutting force signal Sm, state transition equations and observation equations are constructed. The predicted and observed values ​​are fused using a Kalman filter to obtain the corrected load Fi.

[0041] Step 3 outputs balanced cutting loads at each point. However, in actual machining, the nominal depth of cut is affected by various uncertainties related to the machine tool, fixture, cutting tool, and workpiece blank, resulting in certain deviations. This study obtains high-fidelity cutting deformation by fusing measured cutting load values.

[0042] like Figure 5 As shown, the system state equation is constructed for the cutting deformation and cutting force balance process at each sampling point of the global deformation, and it is expressed as follows: ,in: Let k be the calculated cutting force state vector at sampling point k. This is the state transition matrix, and this variable is determined by the DOE fitting parameters for the cutting force modeling. To assign a matrix to the noise, The mean is zero and the variance is The white noise vector.

[0043] Construct the observation equation, which can be expressed as: ,in This is the cutting force observation vector at sampling point k; This is the measurement matrix, which is related to the observation system. If a force sensor is used in the study, then... If it is another type of sensor, it needs to be constructed according to the actual situation. , The mean is zero and the variance is The white noise vector.

[0044] because and It is often difficult to model and solve accurately, making The predicted result is a priori estimate. The estimated value is based on the measurement results. Using the Kalman method, the posterior estimate fits as... , ,in When it is 0, The calculation results have higher reliability. = hour, This indicates that the measurement results have higher reliability.

[0045] Kalman gain ,in For prior results The mean squared error matrix has physical meaning. and For the prior results derived from the model, and For the true result, N is the number of existing samples.

[0046]

[0047]

[0048]

[0049] The uncertainty of the measured cutting load is determined by the observation system such as the force gauge, while the uncertainty of the predicted decoupling is mainly affected by the empirical formula of the cutting force.

[0050] Step 7: Iterate the elastic model again using the corrected load.

[0051] The load Fi, optimized by Kalman filtering, is input into the elastic model, and the iterative calculation in step 3 is repeated to obtain the corrected and accurate deformation value δi. This significantly improves prediction accuracy and robustness, and reduces the cumulative bias caused by initial errors.

[0052] Step 8: Reconstruct the full-field deformation based on spatial correlation

[0053] like Figure 6 As shown, to achieve high-precision prediction of the three-dimensional deformation field of the entire circumferential contour surface under the condition of only sparse measurement data, a reconstruction method based on spatial correlation is proposed. This method uses an XGBoost regression model to construct a spatial reconstruction function. The input is the spatial location and deformation value of sparse sampling points, and the output is the deformation estimate of all surface nodes, thereby realizing the reconstruction prediction of the deformation of the entire field and meeting the industrial-grade accuracy requirements.

[0054] The spatial sampling strategy remains consistent with the "layer-corner" discretization strategy in Step 1 within the overall framework. In Step 1, the workpiece surface was divided into a set of nodes in the polar coordinate system, consisting of "axial layers × circumferential angles," forming the spatial basis coordinate system for model prediction. Building upon this, Step 8 employs a sparse sampling scheme, for example, selecting a subset of nodes of "3 layers × 12 angles" as the source of training samples to significantly reduce the number of measurement points and lower acquisition costs and time overhead.

[0055] This sparse sampling strategy balances spatial representativeness and sampling efficiency, while ensuring good coverage and discriminative power during the model reconstruction training phase. The location of each sampling point ( , Together with its deformation value, it serves as the input to the regression model, and the output is the deformation estimate of the complete discrete node set.

[0056] To enhance the model's adaptability under different deformation modes, extensive trial cutting was conducted. During the training data construction phase, five typical circumferential error morphologies were comprehensively introduced, including standard, inclined, eccentric, conical, and other combined types. For each type, a differentiated initial cutting depth distribution ap was constructed and used as input for elastic modeling and load solving in step 3, obtaining a high-fidelity full-field deformation solution considering actual material parameters and boundary conditions. Furthermore, a subset of points was extracted as model input using the aforementioned sparse strategy to form training samples.

[0057] The XGBoost model has a lightweight overall structure, making it suitable for online deployment. Its inference response time can be controlled within seconds, and it can be used for real-time deformation reconstruction and compensation strategy generation.

[0058] This invention proposes a method for predicting the circumferential profile based on the spatiotemporal correlation between cutting load and deformation field. This method is designed for high-precision in-situ machining of annular thin-walled structural components in high-end manufacturing fields such as aerospace, and has clear practical value and promising prospects for industrial application. Specifically, it is as follows: I. Significant technological advantages, addressing key industrial pain points 1. High-fidelity prediction capability: By integrating elastic mechanism modeling with measured cutting load data, the prediction accuracy of complex circular deformation fields is significantly improved, especially suitable for difficult-to-machine materials such as high-temperature alloys and titanium alloys.

[0059] 2. Fast online response: The use of a lightweight time-varying stiffness model and a fast iteration method can meet the requirements of real-time prediction at the workstation level, which helps to realize closed-loop machining control and online error compensation.

[0060] 3. Low monitoring cost and flexible deployment: The innovative sparse sampling and deformation reconstruction strategy can restore the global deformation field with very few sensors, reducing the difficulty of on-site deployment.

[0061] II. Performance indicators are feasible for engineering implementation.

[0062] III. Conditions for industrialization are mature

[0063] 1. Strong hardware compatibility: The signals relied upon by this invention are data that can be collected by conventional machining centers (such as cutting force, vibration, current, etc.), without the need to introduce expensive sensors, and are compatible with current intelligent machine tool systems.

[0064] 2. Easy to integrate and deploy: The model has a clear structure and a high degree of modularity, and can be quickly deployed on the production line through embedded controllers or edge computing platforms.

[0065] 3. Highly compatible with digital twin / intelligent manufacturing platforms: This invention provides key support for the "prediction" module in the digital twin processing system, and can be integrated with systems such as MES / CAPP / PLC to enhance the overall process control capabilities.

[0066] IV. Wide range of industry demands and significant economic and social benefits.

[0067] There is an urgent need for high-performance manufacturing of annular thin-walled parts in aero-engines, aerospace modules, precision bearings, and gas turbine structural components. The technical solution of this invention has comprehensive benefits such as reducing scrap rate, increasing first-pass yield, and shortening setup time. It can significantly improve the processing yield and unit capacity of enterprises, and play an important role in promoting the localization of high-end manufacturing equipment and the transformation to intelligent manufacturing in my country.

[0068] The preferred embodiments of the present invention have been described in detail above. It should be understood that those skilled in the art can make numerous modifications and variations based on the concept of the present invention without creative effort. Therefore, all technical solutions that can be obtained by those skilled in the art based on the concept of the present invention through logical analysis, reasoning, or limited experimentation on the basis of existing technology should be within the scope of protection defined by the claims.

Claims

1. A circumferential profile prediction method based on the spatiotemporal correlation of cutting forces and deformation fields, characterized in that, The method comprises the following steps: Step 1, spatial discretization of the circumferential profile region of the target workpiece; Step 2, establishing a lightweight structure stiffness matrix; Step 3, construct the elastic mechanism model and solve the δ-F mapping, where represents the resultant force load on the node, represents the elastic deformation variable at the node; Step 4, introducing time correlation of cutting load in the elastic mechanism model; Step 5, constructing the Boolean relationship between the tool path and the workpiece envelope, and extracting the nominal cutting depth as the initial value for solving the elastic mechanism model in step 3; Step 6, fusing the measured process signals for state correction; Step 7, using the corrected load to reiterate the elastic model to obtain the corrected deformation value; Step 8, reconstructing the full-field deformation based on spatial correlation.

2. The method of claim 1, wherein the method is based on the spatiotemporal correlation of the cutting force and the deformation field. The step 1 further comprises: adopting a meshing method to perform surface subdivision on the circumferential profile region of the target workpiece, constructing a discrete node set in a three-dimensional coordinate system, each node representing a local deformation point to be predicted, and adopting a "layer-angle" discretization strategy in a polar coordinate system to perform spatial subdivision on the target circumferential deformation region.

3. The method of claim 1, wherein the method is based on the spatiotemporal correlation of the cutting force and the deformation field. The step 2 further comprises: Step 2.1, based on the geometric envelope relationship between the cutting tool and the target workpiece, spatially partitioning the target workpiece to identify the cutting layer and the non-cutting layer; Step 2.2, using finite element analysis software to model the structure of the spatially partitioned model, and introducing a substructure method to combine the non-cutting layer region into an equivalent super-element to realize the stiffness compression representation of the non-machining region.

4. The method of claim 3, wherein the method is based on the spatiotemporal correlation of the cutting force and the deformation field. The step 2 further comprises: Step 2.3, establishing the overall structure stiffness matrix of the target workpiece in the unprocessed state and the local stiffness matrix collection of all elements in the cutting region, determining the element set of material removal by identifying the intersection of the envelope surface of the cutting tool and the target workpiece at each discrete time step, and using stiffness matrix reduction technology to remove the local stiffness matrix collection of all elements in the cutting region from the overall structure stiffness matrix to obtain the structure stiffness matrix at each discrete time.

5. The method of claim 4, wherein the method is based on the spatiotemporal correlation of the cutting force and the deformation field. The step 3 further comprises: based on the elastic mechanics relation = · , a coupling model between the resultant force load on the node and the elastic deformation variable at the node is established, wherein is the structure stiffness matrix of each discrete time.

6. The method of claim 5, wherein the method is based on the spatiotemporal correlation of the cutting force and the deformation field. The step 3 further comprises constructing a numerical model based on iterative approximation among the force-deformation-cutting depth, the first The iteration process of the first node is as follows: Step 3.1, initialization, setting the initial cutting depth and the initial elastic deformation, and calculating the initial cutting force from the cutting force empirical model; Step 3.2, linear solution, calculating the node deformation using the current load and the structure stiffness matrix at each discrete time; Step 3.3, depth correction, updating the cutting depth according to the deformation result, and recalculating the cutting load accordingly; Step 3.4, convergence judgment, if the difference between the deformation of the first node and the deformation of the second node is less than a given iteration precision, it is considered that the balance is reached, the iteration is ended, otherwise, go back to step 3.2 to continue iteration. Step 3.4, convergence judgment, if the difference between the deformation of the first node and the deformation of the second node is less than a given iteration precision, it is considered that the balance is reached, the iteration is ended, otherwise, go back to step 3.2 to continue iteration.​ 7. The method of claim 6, wherein the method is based on the spatiotemporal correlation of the cutting force and the deformation field. The step 4 further comprises, in the continuous machining process, modeling the duration cutting force data using a hybrid neural network, and the output result is used as the iterative initial value of the initial elastic deformation in step 3.

1.

8. The method of claim 1, wherein the method is based on the spatiotemporal correlation of the cutting force and the deformation field. In the step 6, the measured process signal is the measured value of the cutting load.

9. The method of claim 8, wherein the method is based on the spatiotemporal correlation of the cutting force and the deformation field. The step 6 further comprises: Step 6.1, constructing a system state equation for the cutting deformation and cutting force balance process at each sampling point of the global deformation; Step 6.2, Constructing the observation equation: where, is the cutting force observation vector at sampling point k, is the measurement matrix, is the white noise vector, is the calculated cutting force state vector at sampling point k.

10. The method of claim 1, wherein the method is based on the spatiotemporal correlation of the cutting force and the deformation field. In the step 8, the full-field deformation adopts a sparse sampling scheme, the input is the spatial position and deformation value of the sparse sampling points, and the output is the deformation estimation of the entire surface nodes.