Graphic interactive numerical control machining method and system

Abstract

The invention discloses a graphic interactive numerical control machining method and system. A perceptible digital twin environment integrating geometric, dynamic and thermodynamic multi-source sensor data is constructed; the SLAM and IMU fusion technology is utilized to realize hyperstable alignment of augmented reality and physical space; after graphical processing task definition is carried out in AR, virtual-real fusion rehearsal is carried out through a predictive twin fidelity model, and the stability and risk of the processing process are quantified and predicted by calculating fidelity components of three dimensions of geometry, dynamic and thermodynamics in real time. In rehearsal, the system can find and guide a user to avoid flutter, thermal deformation and other problems in advance. In a physical processing stage, the model continuously drives an adaptive control closed loop, and automatically and smoothly adjusts parameters such as a feed rate according to a predicted fidelity index, or executes safe hovering when a serious risk is predicted.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of intelligent manufacturing and numerical control machining, and specifically relates to a graphical interactive numerical control machining method and system deeply integrating augmented reality (AR), sensor-aware digital twin, and adaptive control strategy based on real-time prediction model. BACKGROUND

[0002] Numerical control (CNC) machining is the core link of modern manufacturing industry. The traditional machining process follows a linear mode of "CAD design → CAM programming → G code generation → machine tool execution". This mode has inherent defects: first, G code is a highly abstract machine language, which is highly dependent on the professional skills and experience of operators, and the code writing, checking and setting process is tedious and prone to errors, which can directly lead to serious accidents such as tool collision and workpiece scrap. Second, there is a huge information barrier between operators and the physical process of machining. Human-computer interaction is limited to a two-dimensional control panel, and it is impossible to intuitively and accurately correlate the machining plan in the digital world with the real-time state of the machine tool and the blank in the physical world. Finally, and most importantly, the simulation in CAM software is carried out in an idealized and disturbance-free digital environment, which cannot take into account dynamic uncertainty factors such as tool wear, uneven material hardness, thermal deformation caused by cutting heat, and system vibration in actual machining. The deviation between "ideal" and "reality" greatly reduces the reliability of simulation, and operators still need to rely on experience and remain vigilant during the machining process, but lack effective, safe and real-time intervention means. In order to improve human-computer interaction, some existing technologies introduce AR technology to superimpose tool paths or design models onto physical machine tools. However, these applications mostly stay at the level of one-way "visualization assistance", and the AR system is only a passive observer, and the virtual information it displays is still idealized data generated offline by CAM, and does not form a closed loop with the real-time physical state of the machine tool. There is still an insurmountable gap between the virtual "seen" and the physical "obtained", and the machining quality and safety problems caused by dynamic uncertainty cannot be solved.

[0003] Therefore, there is an urgent need in the art for a completely new technical solution that can dynamically and in real time assess and bridge the gap between virtual models and physical reality, and upgrade AR from a simple display tool to an intelligent interactive hub capable of prediction, decision-making and adaptive control. SUMMARY

[0004] This section is intended to summarize some aspects of the embodiments of the present application and briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section and the Abstract and Title of the specification to avoid obscuring the purpose of this section, the Abstract and the Title, which cannot be used to limit the scope of the present application.

[0005] The main purpose of the present application is to overcome the fundamental defects in the prior art that the digital model is disconnected with the physical reality and cannot predict and adapt to dynamic uncertainties in the machining process.

[0006] To solve the above technical problems, the present application provides the following technical solutions: a graphic interactive numerical control machining method, comprising the following steps: a) constructing a perceptible digital twin environment: acquiring the geometric data and real-time dynamic state data of physical entities in the working space of the numerical control machine tool, the physical entities including machine tool moving parts, fixtures and workpieces to be machined; the real-time dynamic state data at least includes data streams from machine tool servo encoders, acceleration sensors and temperature sensors; based on the geometric data and real-time dynamic state data, a digital twin model is constructed which is synchronously mapped with the physical entities in terms of geometry, kinematics and physical state; b) establishing a space-time reference: using an augmented reality device, identifying the static structural features of the working space through a simultaneous localization and mapping (SLAM) technology, and fusing the inertial measurement unit (IMU) data of the device, to establish a unified world coordinate system which accurately aligns the digital twin model, the virtual coordinate system of the augmented reality device and the physical coordinate system of the machine tool; c) defining the machining task and generating the initial tool path: through the augmented reality device, selecting the geometric features to be machined on the virtual target model superimposed with the physical workpiece in a graphical interactive manner, and generating the corresponding initial tool path by the system; d) performing a rehearsal verification based on the predictive twin fidelity model: simulating the execution of the initial tool path under the static state of the physical machine tool; during the simulation process, the predictive twin fidelity index F pt (t+Δt) is calculated in real time and rolling for the next time step (Δt), the index is used to quantify the consistency between the digital twin guided machining behavior and its predicted physical result; when the index is lower than the preset verification threshold, the rehearsal is paused and a risk prompt is given at the corresponding position by the augmented reality device; e) adaptive machining execution based on the predictive twin fidelity model: after the rehearsal verification is passed, the machining instructions are sent to the machine tool controller in segments to execute physical machining; during the physical machining process, the predictive twin fidelity index F pt (t+Δt) is continuously calculated in real time, and according to the value of the index, the machining parameters of the machine tool are adaptively adjusted, or a safety hover is executed when the index is lower than the safety threshold.

[0007] As a preferred scheme of the graphic interactive numerical control machining method described in the present application, wherein: the predictive twin fidelity index Fpt (t+Δt) is calculated by the following model:

[0008] F pt (t+Δt) = w g * S g (t) + w d * S d (t) + w t * S t (t)

[0009] Wherein, F pt (t+Δt) is a predictive twin-fidelity index; S g (t) is a geometric fidelity component for representing the deviation degree between the theoretical instruction position and the actual physical position of the tool tip; S d (t) is a dynamic fidelity component for representing the machining vibration stability of the next time step predicted based on the acceleration sensor data; S t (t) is a thermodynamic fidelity component for representing the machining deformation degree caused by thermal effects predicted based on the temperature and power sensor data; w g , w d , w t are preset weight coefficients whose sum is 1.

[0010] As a preferred scheme of the graphic interactive numerical control machining method, the dynamic fidelity component S d (t) is calculated by the following way:

[0011] S d (t) = exp( -k * (A pred (t+Δt) / A crit )² )

[0012] Wherein, A pred (t+Δt) is the vibration amplitude at the next time step t+Δt predicted by using the time series prediction model according to the historical vibration signal sequence collected by the acceleration sensor before the current time t; A crit is the critical vibration amplitude dynamically set according to the current machining process; and k is a sensitivity coefficient.

[0013] As a preferred scheme of the graphic interactive numerical control machining method, the thermodynamic fidelity component S t (t) is calculated by the following way:

[0014] S t (t) = 1 - (ΔL pred (t) / ΔL tol )

[0015] wherein, ΔL pred (t) is the predicted thermal expansion displacement of the critical components based on the readings of temperature sensors and spindle power sensors at the current time t using a real-time thermal deformation surrogate model; ΔL tol is the maximum thermal deformation allowed based on the workpiece tolerance requirement.

[0016] As a preferred scheme of the graphical interactive NC machining method, in step d), the augmented reality device renders the initial tool path in the unified world coordinate system with different visual styles (such as color or transparency) according to the value of the predictive twin fidelity index F pt (t+Δt) to visually indicate the predicted machining risk level.

[0017] As a preferred scheme of the graphical interactive NC machining method, in step e), the adaptive adjustment of the machining parameters of the machine tool comprises: when the value of the predictive twin fidelity index F pt (t+Δt) drops to a preset adjustment interval, the system automatically and smoothly reduces the feed rate and / or adjusts the spindle speed of the machine tool.

[0018] As a preferred scheme of the graphical interactive NC machining method, the adaptive adjustment action performed by the system is presented to the operator through the augmented reality device, and the operator is allowed to confirm, veto or manually fine-tune the adjustment through the graphical interactive interface.

[0019] To solve the above technical problems, the present application also provides the following technical scheme: a graphical interactive NC machining system, characterized in that it comprises: a data acquisition unit: deployed in the working space of the numerical control machine tool, used to acquire the geometric data of the machine tool moving parts, fixtures and workpieces to be machined, and real-time dynamic state data including servo encoder, acceleration and temperature information; an augmented reality device: having a processor, a display, a camera and an inertial measurement unit; a processing controller: in communication connection with the data acquisition unit and the augmented reality device, internally configured to perform the following operations: a) based on the data acquired by the data acquisition unit, a perceivable digital twin model is constructed which is synchronously mapped with the physical entity; b) a simultaneous localization and mapping (SLAM) algorithm is run to fuse the data of the camera and the inertial measurement unit, and a unified world coordinate system is established to align the virtual and physical spaces; c) graphical machining instructions input through the augmented reality device are received, and an initial tool path is generated; d) in the rehearsal verification stage and the physical machining stage, a predictive twin fidelity index F ptan index quantifying the consistency between the digitally-twin-guided machining behavior and its predicted physical outcome; e) based on the index F pt the value of the index F (t+Δt) to give a risk warning in the rehearsal phase and to adaptively adjust the machining parameters of the machine tool or trigger a safety hover in the physical machining phase.

[0020] As a preferred scheme of the graphic interactive numerical control machining system according to the present application, wherein: the model on which the processing controller calculates the predictive twin fidelity index F (t+Δt) is:

[0021] F pt (t+Δt) = w g * S g (t) + w d * S d (t) + w t * S t (t)

[0022] Wherein, S g (t) is a geometric fidelity component, S d (t) is a dynamic fidelity component, S t (t) is a thermodynamic fidelity component, w g , w d , w t are weight coefficients.

[0023] As a preferred scheme of the graphic interactive numerical control machining system according to the present application, wherein: the processing controller is further configured to: utilize a time series prediction model to predict the vibration amplitude A pred (t+Δt) at the next time step according to the historical vibration signal to calculate S d (t); and utilize a real-time thermal deformation replacement model to predict the thermal expansion displacement ΔL pred (t) according to real-time temperature and power data to calculate S t (t).

[0024] The present application provides a graphic interactive numerical control machining method and system, which has the following beneficial effects:

[0025] 1. Greatly improves the safety and reliability of machining: through deep rehearsal taking into account dynamic physical effects and fidelity closed-loop control during machining, it can fundamentally eliminate machining failures caused by complex factors such as vibration, thermal deformation, and overload, and has immeasurable value for machining expensive materials, complex structures, or thin-walled and easily variable parts.

[0026] 2. Realize the adaptive optimization of the processing process: the system can perceive the cutting state in real time and adjust the process parameters autonomously, so that the processing process always runs efficiently and with high quality, thereby stably improving the surface quality and dimensional accuracy of the parts and effectively prolonging the tool life. BRIEF DESCRIPTION OF DRAWINGS

[0027] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced as follows. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor. Among them:

[0028] Figure 1 The method flow chart of the graphical interactive numerical control machining method provided by the present application. DETAILED DESCRIPTION

[0029] In order to make the above-mentioned purposes, features and advantages of the present application more apparent and easy to understand, the specific embodiments of the present application will be described in detail below with reference to the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor should be within the scope of protection of the present application.

[0030] The main purpose of the present application is to overcome the fundamental defects in the prior art that the digital model is disconnected from the physical reality and cannot predict and adapt to dynamic uncertainties in the processing process. Specifically, the present application aims to provide a method and system, which can:

[0031] Before processing, a deep pre-play verification beyond pure geometric collision checking and taking into account dynamic physical effects is performed.

[0032] During processing, the consistency or "fidelity" between virtual instructions and physical execution is quantified in real time.

[0033] Based on the prediction of future processing state, adaptive parameter adjustment is performed in advance, so as to actively avoid risks and optimize processing quality, rather than passively respond to problems that have already occurred.

[0034] Create an intuitive and reliable AR interaction interface, so that operators can make efficient and safe decisions and interventions with full information and system intelligence assistance.

[0035] Specifically, the present application provides a graphical interactive numerical control machining method, comprising the following steps:

[0036] Step 1: Build a multi-level, perceptible digital twin environment

[0037] 1.1. Static and dynamic model construction: Not only the static geometric digital twin of the machine tool, fixture, and blank is established, but more importantly, the multi-source sensor data stream of the servo motor encoder of each axis of the machine tool, the spindle power sensor, the acceleration sensor (which can be installed on the spindle box or workbench), the temperature sensor (installed on the key components), etc. is integrated. This builds a "perceptible" digital twin that can reflect the real-time dynamic state of the physical entity.

[0038] It should be noted that the present application integrates a variety of sensors into the system:

[0039] Servo encoder: Directly read from the numerical control core of the machine tool, providing real-time and high-precision position and speed feedback of each motion axis.

[0040] Acceleration sensor: One or more three-axis MEMS accelerometers are strategically installed in key locations sensitive to vibration, such as the spindle box shell, tool holder, or workbench, to capture high-frequency vibration signals caused by changes in cutting force during machining.

[0041] Temperature sensor: Arranged in key heat-generating components such as spindle bearings, motors, ball screws, and cooling liquid circuits to monitor the thermal state of the system.

[0042] Spindle power / current sensor: By monitoring the power or current consumption of the spindle motor, the size of the real-time cutting force and the generation rate of cutting heat can be indirectly estimated.

[0043] 1.2. Spatiotemporal reference ultra-stable calibration: Abandoning the easily blocked plane Marker, SLAM (Simultaneous Localization and Mapping) technology based on natural features and inertial measurement unit (IMU) fusion is adopted. When the AR device is started for the first time, it will scan and learn the unique three-dimensional structural features of the machine tool work area (such as the edge lines of the machine tool column, the T-shaped slot of the workbench, etc.). At the same time, the IMU data of the AR device is tightly coupled with the visual features. This allows the AR device to maintain sub-millimeter-level, drift-free locking between the virtual overlay layer and the physical world through "feature memory + inertial dead reckoning" even when part of the field of view is blocked or the light changes., laying a solid foundation for subsequent accurate interaction and measurement.

[0044] It needs to be explained: in order to ensure the accurate alignment between the virtual information of AR superposition and the physical entity without jitter and drift, the application adopts advanced SLAM technology based on natural features. When the operator wearing AR equipment (such as Microsoft HoloLens 2 or Magic Leap 2) enters the work area for the first time, the AR system will run an initialization program. The program guides the camera to scan and learn the unique and stable three-dimensional structural feature points (for example, the intersection of a specific corner, bolt hole, T-shaped groove on the machine tool casting, etc.) of the machine tool workspace, and constructs a feature map. At the same time, the IMU (Inertial Measurement Unit) built-in AR equipment provides attitude and acceleration data at a high frequency (such as 200Hz or more). Through fusion algorithms such as Extended Kalman Filter (EKF), the low-frequency but accurate visual feature matching results are tightly coupled with the high-frequency but drifting IMU data. Once the calibration is completed, even if the operator moves, part of the field of view is blocked or the light changes, the system can continuously and stably lock the virtual coordinate system on the physical coordinate system of the machine tool through the combination of "visual repositioning + inertial track prediction", achieving sub-millimeter level tracking accuracy.

[0045] Step two: introduce the Predictive Twin Fidelity (PTF) model. It is not a static model, but a real-time calculation and dynamic evolution evaluation function, which is used to quantify the degree of coincidence between the digital twin guided processing behavior (virtual) and its physical results (reality) in the next time step (Δt).

[0046] The model formula is defined as: F pt (t+Δt) = w g * S g (t) + w d * S d (t) + w t * S t (t)

[0047] Where:

[0048] F pt (t+Δt) : Predictive Twin Fidelity index. It is a dimensionless value between 0 and 1. F pt tends to 1, representing that the system predicts that in the next processing micro-section, the virtual instruction can be highly accurately reproduced in the physical world, the "what you see is what you get" degree is high, and the processing is in a stable and controllable state. F pt tends to 0, representing that there is a large prediction deviation, such as imminent violent vibration, overload or thermal deformation out of control, virtual and reality will be "unhooked", and the risk is extremely high.

[0049] S g (t) : Geometric Fidelity Score. It measures the compliance of the tool and workpiece in the geometric level.

[0050] where, S g (t) = 1 - ||P real (t) - P twin (t)|| / δ max

[0051] P real (t) : Actual physical position vector of the tool tip at time t, fed back by the machine encoder.

[0052] P twin (t) : Theoretical commanded position vector of the tool tip at time t, in the digital twin model.

[0053] ||... || : Euclidean distance operator, i.e., the deviation of two positions.

[0054] δ max : Preset maximum allowed geometric deviation threshold (e.g., 0.05 mm), representing the positioning accuracy limit of the machine or the tolerance band required by the process. When the deviation exceeds this threshold, S g is 0 or negative.

[0055] S d (t) : Dynamic Fidelity Score. It measures the dynamic stability of the machining process, with the core being the predicted vibration.

[0056] where, S d (t) = exp( -k * (A pred (t+Δt) / A crit )² )

[0057] This is a decay model based on the Gaussian function, which is very sensitive to the increase of vibration amplitude. When the predicted vibration amplitude A pred is much smaller than the critical value A crit , the exponential term tends to 0, and S d tends to exp(0)=1. As A pred approaches or exceeds A crit , the absolute value of the exponential term rapidly increases, causing S d to decay exponentially to 0. This can very sensitively reflect the sharp increase of vibration risk.

[0058] A pred(t + Δt): The system utilizes a Kalman filter or similar time-series prediction model to predict the vibration amplitude at the next time step t + Δt based on the vibration signal sequence a(t), a(t - 1),... collected by the acceleration sensor in the past short period of time (e.g., 100 milliseconds). A time-series prediction model (e.g., ARIMA model, LSTM neural network, or simplified Kalman filter) is applied to this sequence to predict the vibration signal amplitude at the future short time step t + Δt. This prediction behavior enables the system to "anticipate" the occurrence of chatter, rather than detecting it after it has occurred.

[0059] A crit : Critical vibration amplitude. This value is dynamically set by the process knowledge base according to the current machining material, tool type, overhang length, and machining method (e.g., cantilever thin-walled part side milling). Exceeding this value, the machining surface will exhibit obvious chatter marks. This is a dynamic threshold, not a fixed value. The system maintains an internal process database that stores simplified data of stability lobe diagrams (Stability Lobe Diagram) for different tools (type, diameter, overhang), materials, and cutting methods (side milling, slotting, etc.). After determining the current machining step, the system queries this database to obtain an A crit value specific to the current specific working condition.

[0060] k: Sensitivity coefficient, used to adjust the severity of the impact of vibration on the fidelity index.

[0061] S t (t): Thermodynamic fidelity component. It measures the impact of thermal deformation on accuracy.

[0062] where, S t (t) = 1 - (ΔL pred (t) / ΔL tol )

[0063] ΔL pred(t) : Real-time simplified thermal deformation model based on finite element analysis (FEA). The model takes spindle power sensor readings (estimated cutting heat), temperature sensor readings (ambient and component temperature) as inputs, and outputs the predicted thermal expansion displacement of key components (e.g. spindle, workpiece) in the current thermal state. The model can be simulated in detail offline using FEA, and online using a response surface model (RSM) or neural network surrogate model for fast computation. The model is trained offline using a large number of detailed FEA simulations. The inputs to the simulation are different spindle speeds (corresponding to heat generation), ambient temperatures, and cooling states, and the outputs are the thermal displacements at key locations (e.g. spindle end face, workpiece key dimensions). After training, a computationally extremely fast response surface model (RSM) or small neural network is generated. During online machining, this surrogate model receives temperature sensor and spindle power sensor (converted to heat generation rate) readings in real time and instantaneously outputs the predicted thermal expansion displacement ΔL pred (t).

[0064] ΔL tol : Maximum thermal deformation allowed by the tolerance requirements of the workpiece. This value is input by the operator according to the current machining of the workpiece drawing tolerance requirements. For example, for a hole with a precision requirement of ±0.02mm, ΔL tol can be set to 0.01mm.

[0065] w g , w d , w t : Weight coefficients. The sum of the three coefficients is 1, and they are dynamically adjusted according to the characteristics of the current machining task. For example:

[0066] When performing large stock rough machining, efficiency is prioritized, and vibration and thermal deformation are not sensitive. At this time, the system can automatically set the weights to {w g =0.7, w d =0.2, w t =0.1}.

[0067] When performing thin-walled part finishing, surface quality (affected greatly by vibration) is the primary goal. The weights can be adjusted to {w g =0.2, w d =0.7, w t =0.1}.

[0068] When machining long and slender precision shaft parts, the problem of dimensional accuracy caused by thermal deformation is most prominent. The weights can be set to {w g =0.2, w d =0.1, w t =0.7}.

[0069] Step 3: AR graphical interaction and rehearsal based on PTF model

[0070] 3.1. Fidelity-driven AR visualization: Instead of a simple semi-transparent model, the operator sees in AR glasses a “living” health status indicator.

[0071] When the predicted F pt index is high (e.g. > 0.9), the AR overlaid tool path and virtual model appear in a stable green color.

[0072] When the F pt index drops (e.g. 0.7-0.9), indicating a potential risk, the path turns into a cautionary yellow color.

[0073] When the F pt index is below the threshold (e.g. < 0.7), indicating that a significant deviation or danger is predicted to occur soon, the path turns into a critical red color, accompanied by flashing and a sound alarm.

[0074] 3.2. Risk-tracing interactive rehearsal: In the pre-machining rehearsal phase, the system not only simulates the geometric motion, but also runs the PTF model completely. When a segment of the rehearsed path turns red, the system pauses. The operator can directly click on the red virtual path, and a window pops up in the AR interface, detailing which component (S g , S d , S t ) score is too low to cause the F pt index to drop, and gives the reason (e.g. “predicted vibration exceeds limit”, “predicted thermal deformation exceeds tolerance”). The operator can adjust the strategy accordingly, for example, if the system alerts a vibration risk, the operator can try to change a shorter tool in AR, or reduce the spindle speed, the system will immediately recalculate and update the F pt index until the path turns green.

[0075] Specifically: when a red path segment appears in the rehearsal, the simulation pauses. The operator clicks on the red path with a gesture, and a diagnostic window pops up in the AR interface, clearly indicating which fidelity component (S g , S d , or S t) is too low, and gives actionable suggestions, e.g., “Dynamic fidelity (Sd) is too low: chatter is predicted. Suggestions: 1. Reduce spindle speed; 2. Reduce depth of cut; 3. Replace with shorter tool.” The operator can directly select one suggestion in the AR interface (e.g., select “Reduce spindle speed” and drag a virtual slider), and the system will immediately recalculate the PTF index for that segment. The operator iteratively adjusts until the whole path is green, and the simulation is validated. This process turns the otherwise empirical process of toolpath optimization into a data-driven, visualized, and interactive process.

[0076] Step Four: PTF Model Driven Adaptive Machining Execution

[0077] 4.1. Fidelity Closed-Loop Command Issuing: Instead of “execute a segment, authorize a segment”, the system adopts a dynamic command flow control based on fidelity thresholds. The controller continuously sends tiny command packets (e.g., motion commands for the next 50 ms) to the machine. Meanwhile, the PTF model makes rolling predictions at a higher frequency (e.g., every 10 ms).

[0078] If the predicted F pt is continuously above the safety threshold, the command flow is sent smoothly.

[0079] If the predicted F pt starts to drop and touches the yellow alert line, the system automatically and smoothly reduces the feed rate and / or adjusts the spindle speed without interrupting the machining, trying to pull the machining state back to the high-fidelity interval. This adaptive adjustment is based on the PTF model’s prediction, not a post-mortem remedy.

[0080] If the F pt drops sharply and breaks the red danger line, the system immediately executes a “safety hover”, evacuates the tool along the pre-computed safety path, and reports the problem source to the operator.

[0081] Specifically, after the simulation is passed, the operator authorizes the start of physical machining. At this time, the processing controller does not simply execute the G code. It adopts a “rolling prediction-authorization execution” closed-loop mode. The controller continuously issues a small segment (e.g., the next 50 ms) of motion commands to the machine. At the same time, the PTF model continuously predicts the machining state that follows at a higher frequency (e.g., every 10 ms).

[0082] Normal execution: If the predicted F pt is continuously green (high fidelity), the command flow is sent smoothly, and the machining proceeds as planned.

[0083] Autonomous fine-tuning: If the predicted F ptinto the yellow zone, indicating that the machining state is trending towards instability. At this point, the system will automatically and smoothly adjust the parameters without human intervention. For example, if S d drops below the red threshold, the system will automatically reduce the feed rate until F pt exponentially rises back to the green zone. This process is transparent to the operator, and the AR interface will prompt "System automatically adjusted the feed rate to 90% to suppress minor vibrations." pt drops below the red threshold, the system will automatically reduce the feed rate until F pt exponentially rises back to the green zone. This process is transparent to the operator, and the AR interface will prompt "System automatically adjusted the feed rate to 90% to suppress minor vibrations."

[0084] Safe Hover: If the predicted F pt drops below the red threshold, the system will immediately execute the highest priority safety action. Instead of simply stopping, it will calculate a collision-free safety path to quickly lift the tool away from the workpiece and then pause the machining. At the same time, the AR interface will report the fault reason to the operator in the most eye-catching way (for example, "Severe overload prediction! Machining has been safely paused").

[0085] 4.2. Operator's Authority Arbitration: When the system makes adaptive adjustments or safe hover, the AR interface will clearly inform the operator: "System detected vibration risk, automatically reduced feed rate from 100% to 85%". The operator has the highest authority, he can accept (Allow), override (Override) or further manually fine-tune (Fine-tune) the system's decision through gestures or voice. This realizes the perfect combination of machine intelligence autonomous optimization and human expert experience supervision.

[0086] Based on the above technical principle explanation, the present application gives the following examples to further illustrate the scheme in the present application:

[0087] Example One

[0088] Fine machining of the side wall of a thin-walled aluminum alloy aviation structural part

[0089] This example aims to demonstrate the application effect of the present application in processing scenarios that are extremely sensitive to vibration.

[0090] 1. Machining scenario and equipment

[0091] Workpiece: 7075-T6 aluminum alloy thin-walled structural part, the area to be machined is a side wall with a height of 80mm and a wall thickness of 2.5mm.

[0092] Machine tool: five-axis linkage CNC machining center.

[0093] Cutting tool: Ø12mm, 4-flute, carbide end mill, tool overhang length 95mm (length-to-diameter ratio close to 8, which is very easy to cause chatter).

[0094] AR device: Microsoft HoloLens 2.

[0095] Sensors: A three-axis accelerometer is installed on the spindle box, a power sensor is equipped on the spindle motor, and a temperature sensor is installed on the machine bed.

[0096] Process objective: Ensure that the surface roughness Ra of the sidewall is ≤0.8μm and there are no obvious chatter marks.

[0097] 2. Preset Parameters and Model Configuration Since this process involves precision machining with extremely high surface quality requirements, and thin-walled parts are highly sensitive to vibration, the weighting coefficient w of the PTF model is set to prioritize dynamic fidelity:

[0098] Weighting coefficient: w g (Geometric) = 0.2, w d (Dynamic) = 0.7, w t (Thermodynamics) = 0.1.

[0099] Dynamic fidelity parameter (S) d ):

[0100] By consulting the process knowledge base, the critical vibration amplitude A that produces visible tool marks under this working condition was determined. crit = 0.3g (g is the acceleration due to gravity).

[0101] Sensitivity coefficient k = 5.0.

[0102] Initial CAM parameters: rotational speed S = 12000 r / min, feed rate F = 2000 mm / min, axial depth of cut ap = 0.5 mm.

[0103] 3. During the pre-performance verification phase, the operator wears a HoloLens 2 and defines the machining task for the sidewall in the AR view. The system generates the toolpath based on the initial CAM parameters and begins the "virtual-real fusion pre-performance".

[0104] Pre-simulation process: The virtual tool moves along the sidewall. At a corner in the middle of the path, due to a sudden change in tool load, the system's background time-series prediction model (LSTM) predicts that the vibration will surge at this point based on the geometry of the area and the cutting parameters.

[0105] Specific data and calculations:

[0106] Before the corner, the predicted vibration amplitude A predApproximately 0.15g.

[0107] S d = exp(-5.0 * (0.15 / 0.3)²) = exp(-5.0 * 0.25) = exp(-1.25) ≈0.286

[0108] Assume that the geometric and thermodynamic fidelities are both ideal (S g =1.0, S t =1.0) at this time, then:

[0109] F pt = 0.2*1.0 + 0.7*0.286 + 0.1*1.0 = 0.2 + 0.2002 + 0.1 = 0.5002

[0110] At this time, F pt index is already low.

[0111] When simulated to the corner, the predicted vibration amplitude A pred (t+Δt) jumps to 0.35g, exceeding the critical value A crit .

[0112] S d (t+Δt) = exp(-5.0 * (0.35 / 0.3)²) = exp(-5.0 * 1.36) = exp(-6.8)≈ 0.0011

[0113] At this time, the dynamic fidelity component drops sharply, almost to zero.

[0114] F pt (t+Δt) = 0.2*1.0 + 0.7*0.0011 + 0.1*1.0 = 0.2 + 0.00077 + 0.1 =0.30077

[0115] AR interaction and analysis:

[0116] Since F pt index (0.30077) is far below the preset verification safety threshold (e.g. 0.7), the rehearsal is paused at this point.

[0117] In the AR view, the virtual tool path at the corner becomes a prominent red color and flashes.

[0118] The operator clicks on the red path, and the AR pops up a diagnostic window: “Risk Warning: Dynamic Fidelity (S_d) is too low (0.0011). Severe chatter is predicted to occur. Recommendations: 1. Reduce the spindle speed to avoid the chatter zone; 2. Reduce the feed rate.”

[0119] Interactive optimization: Operator selects suggestion 1, adjusting the virtual speed from 12000 r / min to 9500 r / min in the AR interface. The system immediately recalculates, and the new A... pred It decreased to 0.2g.

[0120] New S d = exp(-5.0 * (0.2 / 0.3)²) = exp(-2.22) ≈ 0.108

[0121] New F pt = 0.2*1.0 + 0.7*0.108 + 0.1*1.0 = 0.2 + 0.0756 + 0.1 = 0.3756

[0122] F pt The situation has improved somewhat, but the path remains yellow. The operator continues to reduce the feed rate to 1500 mm / min, and the system recalculates. A pred It decreased to 0.12g.

[0123] Final S d = exp(-5.0 * (0.12 / 0.3)²) = exp(-0.8) ≈ 0.449

[0124] Final F pt = 0.2*1.0 + 0.7*0.449 + 0.1*1.0 = 0.2 + 0.3143 + 0.1 = 0.6143

[0125] The path remains yellow, but the risk has been significantly reduced. The operator accepts the plan, the rehearsal continues, and it is successfully completed.

[0126] 4. Adaptive processing execution: After processing begins, the system uses optimized parameters (S=9500, F=1500).

[0127] Process monitoring and adaptive adjustment: When the tool reaches the actual corner area, although the parameters have been optimized, due to the possible slight hardness unevenness within the material, the real-time vibration signal fed back by the acceleration sensor, after being calculated by the predictive model, displays A. pred It shows an upward trend, reaching 0.18g.

[0128] S d = exp(-5.0 * (0.18 / 0.3)²) = exp(-1.8) ≈ 0.165

[0129] F pt= 0.2*1.0 + 0.7*0.165 + 0.1*1.0 = 0.2 + 0.1155 + 0.1 = 0.4155

[0130] F pt The index falls into the preset adaptive adjustment interval (e.g. 0.4-0.6).

[0131] System autonomous intervention: the system triggers the adaptive control logic, automatically and smoothly reduces the feed rate from 1500 mm / min by 15% to 1275 mm / min without notifying the operator.

[0132] Result: after the feed rate is reduced, the cutting load is reduced and the vibration is suppressed. pred Falls back to about 0.1g, pt The index rises above 0.6, and the machining state returns to stability.

[0133] Machining is completed: after the entire side wall is machined, the surface is smooth and there is no chatter mark, completely meeting the requirement of Ra≤0.8μm. This method avoids major risks through pre-rehearsal and eliminates potential quality defects through adaptive control.

[0134] Example Two: Long-time rough machining of deep cavity die steel

[0135] This example aims to demonstrate the application effect of the present application in long-time machining scenarios where the effect of thermal deformation is significant.

[0136] 1. Machining scenario and equipment

[0137] Workpiece: P20 die steel, a cavity with a depth of 150mm needs to be machined.

[0138] Machine tool: three-axis vertical machining center.

[0139] Tool: Ø25mm large helix angle corn milling cutter.

[0140] AR equipment: handheld AR tablet (such as iPad Pro).

[0141] Sensor: spindle power sensor, spindle bearing temperature sensor, workpiece surface non-contact infrared temperature sensor.

[0142] Process target: efficiently remove material while avoiding overcutting or size accuracy exceeding due to thermal expansion of the spindle and workpiece.

[0143] 2. Preset parameters and model configuration Since this working condition is long-time rough machining, heat accumulation is the main contradiction, and it is not sensitive to surface vibration marks. Therefore, the weight of the PTF model is set to focus on thermodynamic fidelity:

[0144] Weighting factor: w g (Geometry) = 0.3, w d (Dynamics) = 0.1, w t (Thermodynamics) = 0.6.

[0145] Thermodynamic fidelity parameter (S t ) :

[0146] The flatness tolerance of the mold cavity bottom surface is required to be 0.05mm, so the maximum allowable thermal deformation ΔL tol = 0.025mm is set.

[0147] Initial CAM parameters: rotation speed S = 1500 r / min, feed rate F = 1800 mm / min, cutting depth ap= 2.0 mm, cutting width ae = 20 mm.

[0148] 3. Rehearsal verification stage The operator uses the AR tablet for rehearsal. The built-in surrogate model in the system starts to predict the heat accumulation according to the cutting parameters.

[0149] Rehearsal process and data calculation:

[0150] After 30 minutes of rehearsal, the thermal deformation surrogate model predicts that the spindle will produce displacement in the Z-axis direction due to thermal elongation according to the continuous high-power cutting input.

[0151] The predicted thermal expansion displacement ΔL pred (t) reaches 0.03mm, which exceeds ΔL tol .

[0152] S_t(t) = 1 - (ΔL pred (t) / ΔL tol ) = 1 - (0.03 / 0.025) = 1 - 1.2 = -0.2

[0153] S t becomes negative, indicating that it has been seriously out of tolerance.

[0154] F pt (t) = 0.3*1.0 + 0.1*1.0 + 0.6*(-0.2) = 0.3 + 0.1 - 0.12 = 0.28

[0155] AR interaction and analysis:

[0156] F pt exponent (0.28) is extremely low, and the rehearsal is paused.

[0157] On the AR tablet, the virtual tool path at the bottom of the cavity turns red.

[0158] Diagnostic information shows: "Risk warning: Thermodynamic fidelity (S t ) too low (-0.2). Predicted Z-axis thermal deformation out of tolerance. Recommendations: 1. Reduce cutting load; 2. Increase mid-process cooling / pause time; 3. Enable machine thermal compensation function (if available)."

[0159] Interactive optimization: The operator does not want to reduce efficiency, and chooses recommendation 2. He sets a periodic cooling macro instruction in the AR interface for the machining program, which lifts the tool to a safe height, stops the spindle and turns on the forced cooling for 2 minutes every 20 minutes of machining. The system re-runs the thermodynamic simulation and shows that after introducing the cooling cycle, the peak of ΔL pred is controlled within 0.018 mm throughout the entire machining process.

[0160] The maximum S t = 1 - (0.018 / 0.025) = 0.28

[0161] At this time, F pt = 0.3*1.0 + 0.1*1.0 + 0.6*0.28 = 0.4 + 0.168 = 0.568

[0162] The path turns yellow, indicating that there is still some risk but it is within a controllable range. The preview passes.

[0163] 4. Adaptive machining execution The machining is performed according to the new program with the added cooling cycle.

[0164] Process monitoring: At the end of the second machining cycle, the workshop environment temperature rises slightly, and the infrared sensor detects that the local temperature of the workpiece is higher than expected.

[0165] Data changes and system responses:

[0166] The thermal deformation surrogate model predicts again that ΔL pred may touch the critical point of ΔL tol before the end of the next cycle.

[0167] The F pt index thus starts to slowly decrease, entering the adaptive adjustment interval.

[0168] System autonomous intervention: The system does not wait until the end of the preset 20 minutes, but triggers the cooling macro instruction 2 minutes in advance when the F pt index touches the yellow warning line (e.g. 0.5). The tool is lifted and forced cooling is started.

[0169] Results: Through this "first step" intelligent intervention, the system successfully controls the peak thermal deformation of the spindle and the workpiece within the tolerance range. Finally, after the cavity machining is completed, the three-coordinate measuring machine detects that the bottom flatness is 0.03mm, which fully meets the design requirements. If the method of the present application is not used, and the traditional continuous processing is used, it is very likely that the bottom will appear 0.05-0.08mm depression or overcut due to thermal deformation.

[0170] Through the above two specific embodiments, it can be clearly seen that the method of the present application can target different processing contradictions (vibration or thermal deformation) by configuring different PTF model weights to achieve accurate risk prediction and targeted problem solving. Its pre-verification and adaptive control capabilities are based on specific, quantifiable data models, thereby transforming the processing process from "experience-based" black box operation to "data-based, predictable, and controllable" transparent and intelligent process.

[0171] In addition, in order to further verify the beneficial effects of the present application, the beneficial effects of Example 1 and Example 2 are quantitatively evaluated:

[0172] Test 1: Thin-walled part processing safety and quality verification

[0173] This test simulates the scenario in Example 1 and aims to verify the risk avoidance and quality improvement effect of the present application in processing vibration-sensitive processing.

[0174] 1. Test setup

[0175] Test object: Process a thin wall (wall thickness 2.5mm, height 80mm) on a 7075-T6 aluminum alloy plate.

[0176] Comparison group (Conventional Method): Use the traditional processing flow. A 3-year experienced operator uses mainstream CAM software (such as Mastercam) to generate G code, and after simple geometric collision checking through the simulation function of the machine tool, directly processes on the machine. The operator adjusts the feed adjustment knob manually according to experience by listening to the sound and looking at the chips.

[0177] Inventive group (Present Invention): Use the method of the present application. A junior operator who has only received 1 day of training on the operation of the present application system uses the present application system for AR interactive pre-visualization, optimization, and adaptive processing.

[0178] Equipment and parameters: The same five-axis machining center, the same batch of tools (Ø12mm long neck end mill) and workpieces were used in both groups. The initial cutting parameters were all recommended by CAM software: S=12000 r / min, F=2000 mm / min, ap=0.5 mm. The PTF model weight settings for the invention group were {w g =0.2, w d =0.7, w t =0.1}.

[0179] Measuring instruments: Coordinate measuring machine (CMM), surface roughness meter, high-frequency vibration analyzer (used to record acceleration sensor data).

[0180] 2. Test process and data recording

[0181] A. Pre-show verification stage (only invention group) The primary operator in the AR pre-show successfully predicted that there was a serious risk of chatter at the corner (F pt index dropped to 0.30). The operator optimized the process parameters according to the AR system prompts: S=9500 r / min, F=1500 mm / min. The entire optimization process took about 3 minutes.

[0182] B. Physical processing stage Both groups carried out physical processing and recorded key data.

[0183] 3. Comparison of test results data

[0184] Table 1: Comparison of safety and processing quality data in Test One

[0185] Comparison item Conventional method Present invention Data analysis and conclusion Processing accident When processing to the corner, severe vibration occurred, producing a loud noise, the tool instantaneously collapsed, and the workpiece side wall was scratched, forcing an emergency stop. Smooth processing throughout, without any abnormal noise or accident. Conclusion 1: Through pre-verification, the invention successfully avoids serious processing accidents (tool damage, workpiece scrap) caused by unreasonable parameters, and the safety is fundamentally improved. Processing time Accident occurred after about 2 minutes, workpiece scrapped, effective processing time was 0. Total processing time 8 minutes 15 seconds (including adaptive speed reduction process). Although the invention group parameters are more conservative, but by avoiding accidents, ensure the successful completion of processing, realize the effective processing output. Maximum vibration amplitude (measured) The vibration amplitude exceeded 1.2g (vibration analyzer recorded) at the moment of the accident. The maximum vibration amplitude throughout the process was controlled within 0.22g. Conclusion 2: The PTF model of the invention is accurate in predicting vibration. Adaptive control effectively maintains the dynamic stability of the processing process at a very high level. Surface roughness (Ra) Workpiece scrapped, unable to measure. After processing, the Ra value of the side wall was between 0.65μm and 0.78μm. Conclusion 3: By precisely suppressing vibration, the invention significantly improves the processing surface quality, and stably meets the finishing requirements. Operator intervention times 0 times (accident occurred without intervention). 0 times (the whole process was adjusted by the system, without human intervention). Conclusion 4: The adaptive ability of the invention frees the operator from the state of needing to be vigilant and ready for manual intervention at any time, reducing labor intensity and dependence on experience.

[0186] Test Two: Verification of the influence of thermal deformation of deep cavity molds and operational complexity

[0187] This test simulates the scenario in Example Two, aiming to verify the effectiveness of the invention in controlling thermal deformation and simplifying the operation process during long-time processing.

[0188] 1. Test setup

[0189] Test object: Process a deep cavity (depth 150mm) on a P20 mold steel.

[0190] Comparison group (Conventional Method): A mold technician with rich experience (10 years of experience) operates. He manually segments and inserts pause instructions (M00) in G code according to his rich experience, in order to manually check and blow cooling during processing. The entire process takes a long time and relies on individual judgment.

[0191] Present Invention: Same primary operator as Test One, using the system of the present invention.

[0192] Equipment and Parameters: Same three-axis machining center, same tool and workpiece batch. Initial cutting parameters S = 1500 r / min, F = 1800 mm / min. The PTF model weight settings for the present invention group are {w g = 0.3, w d = 0.1, w t = 0.6}.

[0193] Measuring Instruments: Three-coordinate measuring machine (CMM), non-contact infrared temperature measuring instrument.

[0194] 2. Test Process and Data Recording

[0195] A. Rehearsal / Programming Stage

[0196] Comparison Group: The master spent about 25 minutes carefully segmenting the program based on experience and manually added 3 cooling pause points.

[0197] Present Invention Group: The primary operator interactively added periodic cooling macro instructions in the AR interface based on the system's prediction of thermal deformation out-of-tolerance risk (F pt < 0.3) during AR rehearsal, which took about 5 minutes.

[0198] B. Physical Machining Stage Both groups started machining, recording total machining time, manual intervention, and key temperature and size data.

[0199] 3. Comparison of Test Results Data

[0200] Table 2: Comparison of Thermal Deformation Control and Operation Efficiency Data in Test Two

[0201] Comparison item Conventional method (expert operation) Present invention (novice operation) Data analysis and conclusion Programming / preparation time 25 minutes (manual segmentation, adding M00 instructions) 5 minutes (AR interactive pre-verification and optimization) Conclusion 5: The invention turns complex, experience-based process planning into a fast, intuitive, and data-supported AR interactive process, with 80% improvement in preparation efficiency. Total processing time 2 hours 48 minutes (including 3 manual checks for cooling, about 5 minutes each time) 2 hours 32 minutes (the system automatically executes 4 cooling cycles, 2 minutes each) Conclusion 6: The intelligent cooling strategy (shorter and more frequent) and automatic execution of the invention, compared with manual judgment and operation, shortens the total processing time by about 9.5%, improving production efficiency. Maximum thermal deformation of cavity bottom (measured) 0.038mm (on the edge of tolerance, close to over tolerance) 0.021mm (stably within the tolerance range) Conclusion 7: The PTF model and adaptive intervention (triggering cooling in advance) of the invention can more accurately control thermal deformation within the target range, with higher processing precision and consistency. Experts also cannot make such accurate real-time judgments. Operator labor intensity / dependence High. Operators need to monitor the whole process, manually pause at predetermined points, operate the air gun, and restart, requiring high concentration. Very low. Operators can leave after the process starts, and the system runs autonomously throughout, only receiving notifications through remote devices when necessary. Conclusion 8: The invention greatly reduces the labor intensity and dependence on high-level skills of operators, enabling novices to complete complex tasks that only experts could handle before, with a huge advantage in labor costs. Peak temperature (workpiece surface) Up to 135°C Through intelligent cooling cycles, the peak temperature is controlled below 105°C. This also indirectly demonstrates the effectiveness of the heat management strategy of the invention, helping to extend tool life and maintain stable workpiece material properties.

[0202] Test Summary

[0203] Based on the above two comparative tests covering typical machining problems, the data conclusively proves the technical effects claimed by the present invention:

[0204] Safety and Reliability: Through predictive rehearsal, catastrophic machining accidents are effectively prevented.

[0205] Machining Quality: Through adaptive control, key influencing factors such as vibration and thermal deformation are quantified and accurately controlled within the target range, resulting in stable and superior product quality compared to traditional methods.

[0206] Efficiency and ease of use: significantly shorten the preparation time of complex process, reduce the uncertainty in the process, free the operator from the heavy and high-pressure monitoring task, and greatly reduce the technical threshold of numerical control machining.

[0207] It should be noted that the above examples are only used to illustrate the technical solutions of the present application and are not limiting. Although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or equivalently replaced without departing from the spirit and scope of the present application, and they should be covered in the scope of the claims of the present application.

Claims

1. A method of graphical interactive numerical control machining, characterized by, Comprising the steps of: a) Constructing a perceptible digital twin environment: acquiring geometric data and real-time dynamic state data of physical entities within the CNC machine workspace, including machine moving parts, fixtures, and workpieces to be machined; real-time dynamic state data at least includes data streams from machine servo encoders, acceleration sensors, and temperature sensors; based on geometric data and real-time dynamic state data, constructing a digital twin model that synchronously maps the physical entities in geometry, kinematics, and physical state; b) Step of establishing a spatiotemporal reference: using an augmented reality device, identifying static structural features of the workspace through simultaneous localization and mapping technology, and fusing inertial measurement unit data of the device, establishing a unified world coordinate system that accurately aligns the digital twin model, the virtual coordinate system of the augmented reality device, and the physical coordinate system of the machine; c) Defining the machining task and generating the initial tool path: through the augmented reality device, selecting the geometric features to be machined on the virtual target model overlaid with the physical workpiece in a graphical interactive manner, and generating the corresponding initial tool path by the system; d) Pre-play validation based on predictive twin fidelity model: simulate the execution of the initial tool path in the physical machine tool at rest; during the simulation, calculate in real time and rolling the predictive twin fidelity index F pt (t+Δt) of the next time step, which quantifies the consistency between the digitally-twin-guided machining behavior and its predicted physical outcome; when the index is lower than a pre-set validation threshold, pause the pre-play and provide a risk prompt at the corresponding location by the augmented reality device; e) Adaptive machining execution based on predictive twin fidelity model: upon pre-play validation, machining instructions are issued to the machine tool controller in segments to execute physical machining; during physical machining, the predictive twin fidelity index F pt (t+Δt) is continuously calculated in real-time and based on the value of the index, adaptive adjustments of the machine tool's machining parameters are made, or a safety hover is executed if the index falls below a safety threshold.

2. The graphical interactive numerically controlled machining method according to claim 1, characterized in that, Predictive twin fidelity index F pt (t+Δt) is calculated from the following model: F pt (t+Δt) = w g * S g (t) + w d * S d (t) + w t * S t (t) wherein F pt (t + Δt) is a predictive twin fidelity index; S g (t) is a geometric fidelity component for characterizing the degree of deviation between the theoretical command position and the actual physical position of the tool tip; S d (t) is a dynamic fidelity component for characterizing the machining vibration stability of the next time step predicted based on the acceleration sensor data; S t (t) is a thermodynamic fidelity component for characterizing the degree of machining deformation caused by thermal effects predicted based on the temperature and power sensor data; w g , w d , w t are preset weight coefficients whose sum is 1. where S g (t) = 1 - ||P real (t) - P twin (t)|| / δ max ; wherein P real (t) is the actual physical position vector of the tool tip at the current time t, as fed back by the machine encoder; P twin (t) is the theoretical commanded position vector of the tool tip at the current time t in the digital twin model; δ max is a pre-set maximum allowed geometric deviation threshold.

3. The graphical interactive numerical control machining method according to claim 2, characterized in that, Dynamic fidelity component S d (t) is calculated by: S d (t) = exp( -k * (A pred (t+Δt) / A crit )² ) wherein A pred (t+Δt) is a vibration amplitude predicted at a next time step t+Δt by using a time series prediction model according to a historical vibration signal sequence collected by the acceleration sensor before a current time t; A crit is a critical vibration amplitude dynamically set according to a current machining process; and k is a sensitivity coefficient.

4. The graphical interactive numerically controlled machining method according to claim 3, characterized in that, thermodynamic fidelity component S t (t) is calculated by: S t (t) = 1 - (ΔL pred (t) / ΔL tol ) where ΔL pred (t) is the predicted thermal expansion displacement of the critical component at the current time t using the real-time thermal distortion surrogate model based on the readings of the temperature sensors and spindle power sensors at the current time t; ΔL tol is the maximum thermal distortion allowed based on the workpiece tolerance requirements.

5. The method of claim 4, wherein: In step d), the augmented reality device renders the initial tool path in the unified world coordinate system with different visual styles according to the value of the computed predictive twin fidelity index F pt (t+Δt) to visually indicate the predicted level of machining risk.

6. The method of claim 5, wherein, In step e), the adaptive adjustment of the machining parameters of the machine tool comprises: when the value of the predictive twin fidelity index F pt (t+Δt) falls to a preset adjustment interval, the system automatically and smoothly reduces the feed rate and / or adjusts the spindle speed of the machine tool.

7. The method of claim 6, wherein: The adaptive adjustment actions performed by the system are presented to the operator through the augmented reality device, and the operator is allowed to confirm, veto, or manually fine-tune the adjustments through a graphical interactive interface.

8. A graphical interactive numerical control machining system, characterized by, Comprising: a data acquisition unit deployed in the CNC machine workspace for collecting geometric data of machine moving parts, fixtures, and workpieces to be machined, as well as real-time dynamic state data including servo encoder, acceleration, and temperature information; an augmented reality device with a processor, a display, a camera, and an inertial measurement unit; a processing controller in communication with the data acquisition unit and the augmented reality device, configured to perform the following operations: a) Based on the data acquired by the data acquisition unit, constructing a perceptible digital twin model that synchronously maps the physical entities; b) Running a simultaneous localization and mapping algorithm, fusing camera and inertial measurement unit data, and establishing a unified world coordinate system to align virtual and physical spaces; c) Receiving graphical machining instructions input through the augmented reality device and generating an initial tool path; d) in the rehearsal verification phase and in the physical machining phase, a predictive twin fidelity index F is calculated in real time, on the fly, which quantifies the consistency between the digitally-twin-guided machining behavior and its predicted physical outcome; pt (t+Δt), which quantifies the consistency between the digitally-twin-guided machining behavior and its predicted physical outcome; e) Based on index F pt the value of (t + Δt), a risk warning is made in the rehearsal phase and the machining parameters of the machine tool are adaptively adjusted or a safety hover is triggered in the physical machining phase.

9. The graphical interactive numerically controlled machining system of claim 8, wherein, The model on which the processing controller calculates the predictive twin fidelity index Fpt(t+Δt) is: F pt (t+Δt) = w g * S g (t) + w d * S d (t) + w t * S t (t) where S g (t) is a geometric fidelity component, S d (t) is a dynamic fidelity component, S t (t) is a thermodynamic fidelity component, w g , w d , w t are weighting coefficients.

10. The graphical interactive numerically controlled machining system of claim 9, wherein, The process controller is further configured to: predict the vibration amplitude A at the next time step using a time series prediction model based on historical vibration signals pred (t+Δt) to calculate S d (t); and predict the thermal expansion displacement amount ΔL pred (t) using a real-time thermal deformation surrogate model based on real-time temperature and power data to calculate S t (t).

Images