A complete-process closed-loop management method for AI-assisted design and NC machining of dentures

Abstract

This application relates to a method, device, equipment, and medium for closed-loop control of the entire process of AI-assisted design and CNC machining of dentures. The method includes: dividing a 3D model of the denture restoration into multiple functional surface regions and constructing a semantic design model based on functional semantic constraints; generating toolpaths based on the semantic design model and attaching semantic mapping information to each tool position point; spatiotemporally aligning and associating machining physical data and semantic mapping information to generate a semantically enhanced machining physical data stream; when any cutting force signal deviates from the normal machining envelope range, marking the corresponding functional surface region as a machining abnormal region and extracting the deviation vector within the machining abnormal region; inputting the deviation vector of the machining abnormal region and the functional semantic constraints into a pre-trained AI inverse correction model to obtain a geometric correction scheme for the machining abnormal region. This method can establish a holographic mapping between design semantics and machining physics.

Description

Technical Field

[0001] This invention belongs to the field of dental prosthesis manufacturing technology, and in particular relates to a closed-loop management method for the entire process of AI-assisted design and CNC machining of dentures. Background Technology

[0002] With the development of technology in the field of dental prosthesis manufacturing, digital denture manufacturing technology has emerged. It realizes the manufacturing of denture entities through computer-aided design (CAD), computer-aided manufacturing (CAM), and CNC machining, and has the characteristics of high processing efficiency, controllable geometric accuracy, and high degree of standardization.

[0003] In traditional techniques, during the CAD design phase, the surface of the denture restoration is divided into geometric regions with clear functional meanings, such as the occlusal contact area. These regions are represented using non-uniform rational B-spline surfaces and assigned specific functional semantic labels. Designers make design decisions based on this semantic information to ensure the accuracy of the shape and spatial position of key functional areas. In the CAM toolpath generation phase, only the geometric information of the CAD model is extracted, and the continuous geometric surface is discretized into toolpath points to generate G-code for CNC machining. During machining, sensors collect abnormal signals such as cutting force fluctuations, tool wear, and machine tool vibration, and attempts are made to use AI systems to reverse-engineer the design based on this feedback data to compensate for machining defects.

[0004] However, when existing CAM systems read CAD models using the above methods, they completely discard the functional semantic labels attached to the surfaces, retaining only pure mathematical and geometric information such as surface equations and control point coordinates. The discretized tool positions only contain spatial positions and tool postures, without carrying any functional surface type information. When an anomaly occurs during machining, although the AI ​​system can collect abnormal signals through sensors and map them to specific tool positions, due to the loss of functional semantics, it cannot determine whether the anomaly occurs in a critical or non-critical area, nor can it clarify the original functional constraints of the abnormal area. This leads to deviations in reverse correction, which may overcompensate non-critical areas or destroy the morphology of critical functional areas, affecting the occlusal accuracy of the denture. Summary of the Invention

[0005] Based on this, it is necessary to provide a closed-loop management method for the entire process of AI-assisted design and CNC machining of dentures, which can establish a holographic mapping bridge between design semantics and processing physics, to address the above-mentioned technical problems.

[0006] Firstly, this application provides a closed-loop management method for the entire process of AI-assisted design and CNC machining of dentures, including:

[0007] S1: Obtain the patient's oral digital impression data, and perform tooth segmentation and occlusal relationship identification based on the oral digital impression data to generate a three-dimensional model of the prosthesis;

[0008] S2: Based on prior anatomical knowledge, multiple functional surface regions are divided on the three-dimensional model of the prosthesis, and a semantic design model based on functional semantic constraints is constructed for the functional surface regions; wherein, the functional semantic constraints include the geometric error tolerance range corresponding to each functional surface region and the correction constraint function used to limit the feasible domain of geometric correction.

[0009] S3: Generate toolpaths covering each functional surface region based on the semantic design model based on functional semantic constraints, and attach semantic mapping information to each tool point of each toolpath; the semantic mapping information is used to characterize the attribution association between the tool point and the corresponding functional surface region and functional semantic constraints respectively;

[0010] S4: Acquire the machining physical data collected in real time during CNC machining, and perform spatiotemporal alignment and association between the machining physical data and the semantic mapping information of the corresponding tool position points to generate a semantically enhanced machining physical data stream; wherein, the machining physical data includes cutting force signals;

[0011] S5: Perform anomaly detection on semantically enhanced machining physical data stream. When the cutting force signal in any functional surface region deviates from the normal machining envelope range, mark the corresponding functional surface region as a machining anomaly region and extract the deviation vector of the machining physical data in the machining anomaly region.

[0012] S6: Input the deviation vector and functional semantic constraints of the processing anomaly region into the pre-trained AI inverse correction model. The AI ​​inverse correction model solves the geometric correction feasible region within the limit of the correction constraint function to obtain the geometric correction scheme for the processing anomaly region.

[0013] S7: Apply the geometric correction scheme to the corresponding functional surface region in the semantic design model based on functional semantic constraints to obtain the corrected semantic design model based on functional semantic constraints. Then, based on the corrected semantic design model based on functional semantic constraints, execute S3 to S6 repeatedly until the deviation vectors in all processing abnormal regions fall within the geometric error tolerance range of the corresponding functional surface region.

[0014] Secondly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of any of the above-mentioned closed-loop control methods for the entire process of AI-assisted design and CNC machining of dentures.

[0015] Thirdly, this application also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of any of the above-mentioned AI-assisted design-CNC machining closed-loop control methods for dentures.

[0016] The aforementioned AI-assisted design-CNC machining closed-loop management method for the entire process of dentures generates a 3D model of the denture prosthesis from the patient's oral digital impression data. Based on prior anatomical knowledge, functional surface regions with clearly defined functions are delineated on the model surface. A semantic design model carrying geometric error tolerance ranges and correction constraint functions is constructed, embedding functional protection genes into the design model. When generating toolpaths, the semantic constraint information of the functional surface regions is attached to each tool position in the form of semantic mapping information, opening up the transmission channel from CAD design semantics to the CAM machining end. During CNC machining, by spatiotemporally aligning and associating real-time acquired cutting force signals and other machining physical data with the semantic mapping information of the corresponding tool positions, a semantically enhanced machining physical data stream carrying functional semantic labels is generated. This ensures that the detection of machining anomalies is no longer limited to "high force value at a certain spatial position". It goes beyond a superficial understanding of "abnormalities" to accurately determine which functional area an anomaly occurs in and the level of its functional importance. When a cutting force signal deviates from the normal machining envelope, the deviation vector of the abnormal area is extracted and input into the AI ​​reverse correction model along with the corresponding functional semantic constraints. The correction scheme is solved within the geometrically feasible domain defined by the correction constraint function, thus ensuring that the correction behavior always complies with functional constraints such as the occlusal contact surface not penetrating the opposing tooth and the marginal closure surface not deviating from the termination line. After applying the correction scheme to the semantic design model, the path generation, machining monitoring, and AI correction steps are executed in a loop until the deviation vectors of all abnormal areas fall within the tolerance range. This achieves a holographic mapping and closed-loop iteration between the design functional semantics and the machining physical feedback, and completes the intelligent and precise correction of machining defects while protecting the anatomical integrity of the key functional surfaces. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments or related technologies of this application, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart illustrating the closed-loop control method for the entire process of AI-assisted design and CNC machining of dentures according to the present invention.

[0019] Figure 2 This is a flowchart illustrating the steps of step S3. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0021] In one embodiment, such as Figure 1 As shown, a closed-loop management method for the entire process of AI-assisted design and CNC machining of dentures is provided. This embodiment illustrates the application of this method to a dental prosthesis manufacturing terminal. It is understood that this method can also be applied to a dental prosthesis manufacturing server, and further to a dental prosthesis manufacturing system including both a dental prosthesis manufacturing terminal and a dental prosthesis manufacturing server, and is implemented through the interaction between the dental prosthesis manufacturing terminal and the dental prosthesis manufacturing server. In this embodiment, the method includes the following steps:

[0022] S1: Obtain the patient's oral digital impression data, and perform tooth segmentation and occlusal relationship identification based on the oral digital impression data to generate a three-dimensional model of the prosthesis.

[0023] Indicatively, the dental prosthesis manufacturing terminal can acquire digital impression data of the patient's oral cavity using devices such as intraoral scanners and desktop scanners. This digital impression data can be in various formats, including point cloud data and polygonal mesh data. The dental prosthesis manufacturing terminal can preprocess the acquired digital impression data, including operations such as denoising, smoothing, and registration, to eliminate noise and errors generated during data acquisition and ensure data accuracy. To achieve tooth segmentation, the dental prosthesis manufacturing terminal can employ a deep learning-based image segmentation algorithm to construct a tooth segmentation model. The loss function of the tooth segmentation model can be the cross-entropy loss function, the expression of which is as follows: ,in, This represents the cross-entropy loss value. This represents the total number of pixels in the digital dental impression data. Indicates the first The true label of a pixel, when that pixel belongs to the tooth region. When it is not in the tooth area , This indicates that the tooth segmentation model predicts the first tooth. The probability that a pixel belongs to a tooth region. The dental prosthesis manufacturing terminal trains a tooth segmentation model using this loss function, and uses the trained model to segment preprocessed digital dental impression data to obtain individual tooth region data.

[0024] Furthermore, during the occlusal relationship identification process, the dental prosthesis manufacturing terminal can use the Iterative Closest Point (ICP) algorithm for registering opposing teeth, the core formula of which is: ,in, Represents the rotation matrix. Represents the translation vector. Indicates the number of registration point pairs. The first tooth in the maxillary tooth region Coordinates of a point, The first tooth in the mandibular region Coordinates of a point, This represents the Euclidean norm. The dental prosthesis manufacturing terminal uses ICP (Internal Prosthetic Processing) to solve for the relative spatial relationship between the upper and lower jaw teeth, i.e., the occlusal relationship. Based on the segmented tooth region data and the identified occlusal relationship, the dental prosthesis manufacturing terminal can use the Poisson reconstruction algorithm to generate a 3D model of the prosthesis. The Poisson reconstruction algorithm fits the point cloud data by constructing an implicit function, which satisfies... ,in, Represents the Laplace operator. The implicit function is represented by this partial differential equation. By solving this equation, the three-dimensional surface model of the prosthesis is obtained, thus completing the generation of the three-dimensional model of the prosthesis.

[0025] S2: Based on prior anatomical knowledge, multiple functional surface regions are divided on the three-dimensional model of the prosthesis, and a semantic design model based on functional semantic constraints is constructed for each functional surface region. The functional semantic constraints include the geometric error tolerance range corresponding to each functional surface region and the correction constraint function used to limit the feasible domain of geometric correction.

[0026] Furthermore, the dental prosthesis manufacturing terminal can determine the key functional surface types of the prosthesis based on prior knowledge of oral anatomy, including but not limited to the occlusal contact surface area, the marginal closure surface area, and the proximal surface area. Optionally, the dental prosthesis manufacturing terminal can use a region growing algorithm to divide the functional surface regions of the three-dimensional model of the prosthesis. The core formula of the region growing algorithm is as follows: ,in, Points on the three-dimensional model of the denture restoration and points The Euclidean distance between them Point spatial coordinates, Point The spatial coordinates of the prosthesis are defined. The dental prosthesis manufacturing terminal can set a distance threshold, starting from a preset seed point, and include adjacent points less than the threshold that meet anatomical characteristics into the same functional surface region, gradually expanding to complete the division of all functional surface regions. After division, the dental prosthesis manufacturing terminal constructs a semantic design model based on functional semantic constraints for each functional surface region. The semantic design model can adopt a multi-channel data structure, integrating geometric representation data and functional semantic constraint data. The geometric error tolerance range in the functional semantic constraints can be defined by an error evaluation function, the expression of which is: ,in, Indicates the range of geometric error tolerance. This represents the normal error tolerance value. This represents the tangential error tolerance value, used to limit the allowable deviation range of the geometric shape of the functional surface region. The correction constraint function is used to limit the feasible region for geometric correction. For the meshing contact surface region, its correction constraint function can be expressed as: ,in, This represents the modified constraint function. This represents the spatial coordinate vector of any point on the occlusal contact surface area. The point set representing the surface of the opposing teeth. Let represent the spatial coordinates of any point on the opposing tooth surface. This correction constraint function indicates that the corrected occlusal contact area must not penetrate the opposing tooth surface. For the marginal closed surface area, the correction constraint function can be expressed as: ,in, This represents the modified constraint function. This represents the thickness at any point on the edge-closed region. The minimum allowable thickness of the edge-closed surface region is defined by this correction constraint function, which indicates that the corrected thickness of the edge-closed surface region must not be less than the minimum allowable thickness. The dental prosthesis manufacturing terminal can aggregate the geometric representation data and functional semantic constraint data of all functional surface regions to construct a unified semantic design model based on functional semantic constraints.

[0027] S3: Generate toolpaths covering each functional surface region based on the semantic design model based on functional semantic constraints, and attach semantic mapping information to each tool point on each toolpath; the semantic mapping information is used to characterize the attribution relationship between the tool point and the corresponding functional surface region and functional semantic constraints.

[0028] Optionally, the dental prosthesis manufacturing terminal can select a suitable toolpath generation algorithm, including but not limited to the isoparametric method and adaptive sampling method, based on a semantic design model with functional semantic constraints, combined with the material properties and machining accuracy requirements of the prosthesis, to generate toolpaths covering all functional surface regions. For example, the formula for calculating the tool position coordinates using the isoparametric method is: ,in, The spatial coordinate vector representing the tool position point. and These represent the NURBS surface in direction and Number of control points in direction and These represent the NURBS surfaces in... direction and The order of direction and They represent direction and B-spline basis functions in the direction, These represent the coordinates of the control points on the NURBS surface. The dental prosthesis fabrication terminal can adjust these parameters. and The step size of the value controls the density of the tool position points, ensuring that the tool path can completely cover each functional surface area while meeting the machining accuracy requirements.

[0029] Furthermore, during the toolpath generation process, the dental prosthesis manufacturing terminal can attach semantic mapping information to each tool position. This semantic mapping information can be obtained through spatial distance calculation, and the formula is as follows: ,in, Indicates the first The minimum distance from each tool point to the center point of each functional surface region Indicates the first Spatial coordinate vectors of each tool position point Indicates the total number of functional surface regions. Indicates the first The center point coordinate vector of each functional surface region. The dental prosthesis manufacturing terminal determines the functional surface region corresponding to the minimum distance as the functional surface region to which the cutting point belongs. Then, it extracts the functional semantic constraint information corresponding to the functional surface region, and uses information such as the functional surface region identifier and functional semantic constraint parameters as semantic mapping information, which is then attached to the corresponding cutting point. The semantic mapping information can be stored in the form of key-value pairs and written into the toolpath file along with the spatial coordinates of the cutting point and the tool posture, ensuring that each cutting point can be clearly associated with the corresponding functional surface region and functional semantic constraints.

[0030] S4: Acquire the machining physical data collected in real time during CNC machining, and perform spatiotemporal alignment and association between the machining physical data and the semantic mapping information of the corresponding tool position points to generate a semantically enhanced machining physical data stream; wherein, the machining physical data includes cutting force signals.

[0031] Specifically, the dental prosthesis manufacturing terminal can establish a communication connection with the sensors of CNC machining equipment to acquire real-time machining physical data during CNC machining. This data includes not only cutting force signals but also vibration acceleration signals, spindle power signals, and acoustic emission signals. The dental prosthesis manufacturing terminal can perform synchronous preprocessing on the acquired machining physical data, including filtering, amplification, and normalization, to remove noise interference and ensure data validity. The normalization of the machining physical data can use a linear normalization formula, the expression of which is: ,in, This represents the normalized processing physical data value. Represents the raw measurement values ​​of the physical data of the process. The minimum measurement value for processing physical data, This represents the maximum measured value of the physical data being processed.

[0032] Furthermore, to achieve spatiotemporal alignment and association between processing physical data and semantic mapping information of tool points, the dental prosthesis manufacturing terminal can employ a timestamp synchronization algorithm, namely... ,in, This represents the time difference between the acquisition time of machining physical data and the tool position running time. Indicates the timestamp for the acquisition of physical data during processing. This indicates the timestamp of the cutting point operation. The dental prosthesis manufacturing terminal can be set with a time difference threshold; when... When the value is less than a threshold, the processing physical data is determined to match the corresponding tool position, and the semantic mapping information of that tool position is appended to the processing physical data record. Simultaneously, the dental prosthesis manufacturing terminal can construct a semantic mapping index table, the index construction formula of which is... ,in, Represents the semantic mapping index value. This represents the index mapping function. The spatial coordinate vector representing the tool position point. This represents the tool orientation vector at the tool position point. Through this semantic mapping index table, the dental prosthesis manufacturing terminal can quickly retrieve the semantic mapping information of the tool position point corresponding to each machining physical data point, achieving efficient association between the two and ultimately generating a semantically enhanced machining physical data stream containing machining physical data, tool position point information, and semantic mapping information.

[0033] S5: Perform anomaly detection on semantically enhanced machining physical data stream. When the cutting force signal in any functional surface region deviates from the normal machining envelope range, the corresponding functional surface region is marked as a machining anomaly region, and the deviation vector of the machining physical data in the machining anomaly region is extracted.

[0034] Indicatively, the dental prosthesis manufacturing terminal can perform anomaly detection on the semantically enhanced processing physical data stream, focusing on the detection of cutting force signals. It can also combine other processing physical data such as vibration signals and acoustic emission signals to improve detection accuracy. Specifically, the dental prosthesis manufacturing terminal can construct a normal processing envelope based on cutting mechanics theory, with the core formula being: ,in, Indicates the main cutting force. This represents the unit cutting force coefficient of the material, which can be calibrated according to the type of denture material. Indicates the depth of cut. This represents the feed per tooth. The upper and lower boundaries of the normal machining envelope can be calculated using the main cutting force and the tolerance factor, i.e. , ,in, This represents the upper bound of the normal processing envelope. This represents the lower bound of the normal processing envelope. This represents the tolerance factor, used to cover minor fluctuations during normal processing.

[0035] For example, the dental prosthesis manufacturing terminal can use a sliding window method to monitor the cutting force signal in real time. The length of the sliding window can be adjusted according to the processing feed rate, and the statistical formula for the signal within the window is as follows: ,in, This represents the average value of the cutting force signal within the window. Indicates the length of the sliding window. Indicates the first in the window The measured value of each cutting force signal. When the average or instantaneous value of the cutting force signal within the window deviates from the normal machining envelope range, i.e. or At this time, the dental prosthesis manufacturing terminal marks the functional surface area corresponding to the cutting force signal as a processing abnormality area. Further, the dental prosthesis manufacturing terminal extracts the deviation vector of all processing physical data within the processing abnormality area. The formula for calculating the deviation vector is... ,in, This represents the deviation vector of the physical data of the processing. This represents a vector of actual measured values ​​of the physical data being processed. This represents the standard value vector of physical data processed under normal processing conditions. This deviation vector can accurately reflect the degree and direction of processing anomalies.

[0036] S6: Input the deviation vector of the processing anomaly region and the functional semantic constraints into the pre-trained AI inverse correction model. The AI ​​inverse correction model solves within the geometric correction feasible region defined by the correction constraint function to obtain the geometric correction scheme for the processing anomaly region.

[0037] Optionally, the dental prosthesis manufacturing terminal can input the deviation vector of the processing abnormal area and the corresponding functional semantic constraints of that area into a pre-trained AI inverse correction model. This AI inverse correction model can adopt a deep neural network model with an encoder-decoder structure, and its core loss function is... In the formula, This represents the total loss value of the model. and Let represent the weight coefficients of the geometric prediction error term and the constraint violation penalty term, respectively. The optimal combination can be determined through cross-validation. The geometric prediction error term is expressed as follows: ,in, This indicates the number of control points in the functional area. Indicates the first The true correction vector of each control point The model predicts the first The correction vector for each control point The expression for the constraint violation penalty term is: ,in, Indicates the number of constraint checkpoints. Indicates the first The corrected constraint function value corresponding to each checkpoint.

[0038] Specifically, during the inference process, the AI ​​inverse correction model can extract multimodal features from the input bias vector and functional semantic constraints. The bias vector feature extraction can be achieved through a one-dimensional convolutional layer, while the geometric error tolerance range and correction constraint function in the functional semantic constraints can be converted into dense vectors through an embedding layer. The extracted multimodal features are then fused and input to the decoder. The initial geometric correction scheme output by the decoder needs to satisfy the geometric correction feasible region defined by the correction constraint function. The dental prosthesis manufacturing terminal can optimize the initial correction scheme using a constraint projection algorithm, the optimization formula of which is as follows: The constraints are ,in, This represents the optimized geometric correction scheme. This represents the initial geometric correction scheme output by the model. This represents the original control point coordinate vector of the functional area. This represents the correction constraint function. By solving this constrained optimization problem, the dental prosthesis manufacturing terminal obtains a geometric correction scheme for the processing anomaly region that satisfies the functional semantic constraints. This geometric correction scheme can be output in the form of a control point correction vector.

[0039] S7: Apply the geometric correction scheme to the corresponding functional surface region in the semantic design model based on functional semantic constraints to obtain the corrected semantic design model based on functional semantic constraints. Then, based on the corrected semantic design model based on functional semantic constraints, execute S3 to S6 repeatedly until the deviation vectors in all processing abnormal regions fall within the geometric error tolerance range of the corresponding functional surface region.

[0040] Furthermore, the dental prosthesis manufacturing terminal can locate the corresponding functional surface region in the semantic design model based on functional semantic constraints, based on the identification information of the processing abnormal area. The geometric correction scheme output by the AI ​​reverse correction model is then applied to this functional surface region. The application of the geometric correction scheme mainly involves adjusting the NURBS surface control points of the functional surface region, and the adjustment formula is as follows: ,in, Indicates the corrected number The coordinate vector of each control point Indicates the first [number] before the correction. The coordinate vector of each control point In the geometric correction scheme, the first The correction vector for each control point. After adjustment, the dental prosthesis manufacturing terminal can reconstruct the NURBS surface of this functional area, using the following formula: ,in, This represents the spatial coordinate vector of the modified NURBS surface. This represents the corrected control point coordinates; the remaining parameters are consistent with the NURBS surface parameters used in the toolpath generation process. To ensure the integrity and continuity of the 3D model of the prosthesis, the dental prosthesis manufacturing terminal can perform a continuity check on the corrected functional surface areas and adjacent functional surface areas. The check formula is as follows: ,in, This represents the normal vector of the modified functional surface region at its boundary. This represents the normal vector of adjacent functional surface regions at the boundary. The threshold value represents the normal vector deviation. When the deviation exceeds this threshold, the dental prosthesis manufacturing terminal can adjust the coordinates of control points near the boundaries of adjacent regions to eliminate surface discontinuities, thus obtaining a corrected semantic design model based on functional semantic constraints. Specifically, based on the corrected semantic design model, the dental prosthesis manufacturing terminal iteratively executes steps S3 to S6, regenerating the toolpath, acquiring machining physical data, performing anomaly detection, and geometric correction. The termination condition for the iterative iteration is... ,in, This represents the norm of the deviation vector within the abnormal processing region. This represents the geometric error tolerance range for the corresponding functional surface region. When the deviation vectors in all processing abnormal regions fall within the geometric error tolerance range of the corresponding functional surface region, the dental prosthesis manufacturing terminal terminates the loop iteration and outputs the final prosthesis design and processing scheme.

[0041] In the aforementioned closed-loop management method for the entire process of AI-assisted design and CNC machining of dentures, a semantic design model with functional semantic constraints is constructed. Semantic mapping information between the corresponding functional surface region and the functional semantic constraints is added to the tool path's tool position point. The physical data of CNC machining and the semantic mapping information are spatiotemporally aligned and correlated to generate a semantically enhanced machining physical data stream. Anomalies are detected in this data stream, and abnormal machining areas are marked and deviation vectors are extracted. The deviation vectors and functional semantic constraints are input into a pre-trained AI inverse correction model. A geometric correction scheme is obtained within the feasible domain defined by the correction constraint function. The correction scheme is applied to the semantic design model and iterated until the deviation meets the geometric error tolerance range. This fundamentally solves the semantic discontinuity problem in the CAD-CAM data conversion process, achieving accurate semantic judgment of machining anomalies and constraint-aware geometric correction. It ensures the morphological integrity and functional adaptability of key functional areas of the denture restoration, effectively improving the accuracy, rationality, and overall efficiency of digital denture machining.

[0042] In one embodiment, a semantic design model based on functional semantic constraints is constructed for each functional surface region, including:

[0043] S21: NURBS surface representation of denture restoration based on 3D model of denture restoration. By using the boundary curves of each functional surface region as trimming curves, sub-surface patches of each functional surface region are extracted, and the control point coordinate matrix, node vector sequence, weight sequence and surface order of each sub-surface patch are saved as geometric representation data of the functional surface region.

[0044] Indicatively, the dental prosthesis manufacturing terminal can convert a 3D model of a dental prosthesis into a NURBS surface representation. The general expression for a NURBS surface of a dental prosthesis is: ,in, This represents the spatial coordinate vector of any point on the NURBS surface of the denture restoration. and These represent the NURBS surfaces in... direction and Number of control points in direction Indicates the first Line number The weights of column control points, Indicates the first Line number The spatial coordinate vector of the column control points, and They represent direction Rank, direction The B-spline basis functions of order [order missing]. The dental prosthesis fabrication terminal can obtain the boundary curves of each functional surface region using a boundary extraction algorithm. The NURBS representation of the boundary curves is... ,in, This represents the spatial coordinate vector of any point on the boundary curve. This indicates the number of control points on the boundary curve. Indicates the boundary curve of the first The weights of each control point Indicates the boundary curve of the first Spatial coordinate vectors of control points express B-order spline basis functions, This represents the parameter variable. The dental prosthesis manufacturing terminal can use this boundary curve as a trimming curve to trim the NURBS surface of the prosthesis. The core formula for the trimming operation is... ,in, This represents the functional surface region sub-surface patch obtained after trimming. Represents the pruning feature function, when When located within the area enclosed by the boundary curve ,otherwise After trimming, the dental prosthesis manufacturing terminal extracts the control point coordinate matrix, node vector sequence, weight sequence, and surface order corresponding to the subsurface patch, saving them as the geometric representation data of the functional surface region. The control point coordinate matrix is... The node vector sequence includes Direction node vector and Direction node vector The weight sequence is The order of the surface is .

[0045] S22: Create semantic label data for each functional surface region; the semantic label data includes functional surface type identifier, functional priority weight, normal error tolerance value, tangential error tolerance value, and correction constraint function; wherein, if the functional surface type identifier is an occlusal contact surface type, the correction constraint function of the functional surface region includes a first constraint term and a second constraint term; if the functional surface type identifier is a marginal closure surface type, the correction constraint function of the functional surface region includes a third constraint term and a fourth constraint term; the first constraint term is used to constrain the corrected sub-surface not to penetrate the surface of the opposing tooth in the contact area; the second constraint term is used to constrain the average curvature change of the corrected sub-surface not to exceed a preset curvature change threshold in the central contact sub-region; the third constraint term is used to constrain the corrected sub-surface not to deviate from the surface of the abutment tooth preparation termination line by more than a preset distance threshold in the marginal closure region; the fourth constraint term is used to constrain the local thickness of the corrected sub-surface not to be less than a preset minimum thickness value in the marginal closure region.

[0046] Optionally, the dental prosthesis manufacturing terminal can create semantic label data for each functional surface area. This semantic label data uses an extensible data structure, allowing for flexible field expansion based on different functional surface types. Functional surface type identifiers can be represented by enumerated values ​​to uniquely identify the type of functional surface area, such as occlusal contact surface type, marginal closure surface type, etc. Functional priority weights can use normalized values, calculated using the following formula: ,in, Indicates the priority weight of functions. This represents the original priority score of the functional area. This represents the maximum original priority score for all functional surface regions; a higher value indicates that the region is more critical to the denture's function. The normal error tolerance value and the tangential error tolerance value together constitute the geometric error tolerance range, used to constrain the allowable deviations in the normal and tangential directions of the subsurface patches, respectively. For example, when the functional surface type identifier is an occlusal contact surface type, the expression for the first constraint term is: ,in, Indicates the first constraint term. This represents the spatial coordinate vector of any point within the contact area of ​​the corrected subsurface patch. The point set representing the surface of the opposing teeth. Let represent the spatial coordinate vector of any point on the opposing tooth surface. The first constraint ensures that the modified subsurface does not penetrate the opposing tooth surface. The expression for the second constraint is: ,in, This indicates the second constraint term. This indicates the number of sampling points within the central contact sub-region. This indicates that the corrected subsurface patch is in the first position. The curvature value of each sampling point Indicates the subsurface patch before correction at the 1st... The curvature value of each sampling point This represents the preset curvature change threshold. When the functional surface type identifier is an edge-closed surface type, the expression for the third constraint term is: ,in, This indicates the third constraint term. This represents the spatial coordinate vector of any point within the closed region of the edge of the corrected subsurface patch. The set of points representing the surface at the termination line of the prepared abutment tooth. This represents the spatial coordinate vector of any point on the surface of the termination line of the prepared abutment tooth. Represents the preset distance threshold; the expression for the fourth constraint term is: ,in, This indicates the fourth constraint term. The local thickness at any point within the closed region of the edge of the corrected subsurface patch is represented by the following formula: In the formula, This represents the point coordinate vector corresponding to the abutment tooth. express Vectors and subsurfaces in The angle between the point normal vectors, The minimum thickness value is set.

[0047] S23: Combine the geometric representation data and semantic label data of each functional area into a dual-channel data structure.

[0048] Furthermore, the dental prosthesis manufacturing terminal can combine the geometric representation data and semantic label data of each functional surface region into a dual-channel data structure. This dual-channel data structure enables the organic integration of geometric and semantic information, facilitating subsequent toolpath generation and AI reverse engineering. The dual-channel data structure is represented as follows: ,in, Indicates the first Dual-channel data structure for each functional area This represents the geometric representation data of the functional surface region. Semantic label data representing this functional area. Geometric representation data. The expression is ,in, Indicates the first The coordinate matrix of control points for each functional surface region sub-surface patch. and These represent the subsurface patches respectively. direction and A sequence of node vectors in a direction. This represents the weight sequence of the subsurface patch. and These respectively represent the subsurface patch in direction and The order of the surface in the direction. Semantic tag data. The expression is ,in, Indicates the first Functional surface type identifier for each functional surface region. This indicates the functional priority weight of the functional area. and These represent the normal error tolerance value and the tangential error tolerance value of the functional surface region, respectively. This represents the set of correction constraint functions for the functional surface region. Optionally, the dental prosthesis manufacturing terminal can store this dual-channel data structure in a structured data format to ensure a one-to-one correspondence between geometric representation data and semantic label data. Simultaneously, the data can be compressed using the following compression formula: ,in, This represents the compressed dual-channel data structure. This represents a data compression function. This indicates the compression factor, which facilitates data storage and transmission.

[0049] S24: Aggregate the dual-channel data structures of all functional surface regions into a semantic design model based on functional semantic constraints, and establish a topological association graph structure in the semantic design model based on functional semantic constraints to represent the interlocking association relationship and adjacent surface association relationship between each functional surface region.

[0050] Specifically, the dental prosthesis manufacturing terminal can aggregate the dual-channel data structures of all functional areas to form a unified semantic design model based on functional semantic constraints. The expression of the semantic design model is as follows: ,in, This represents a semantic design model based on functional semantic constraints. Indicates the total number of functional surface regions. ( This represents the dual-channel data structure for each functional area. A topological relationship graph structure representing the connections between functional regions. It can be represented as ,in, Represents the set of nodes in a topological graph. This represents the set of edges in a topological graph. The set of nodes... Each node Dual-channel data structure corresponding to a functional area Node attributes include the region identifier and the function face type identifier for the function face region. Edge set. This includes occlusal association edges and proximal association edges. Occlusal association edges are used to connect the occlusal contact area of ​​a prosthesis with its corresponding opposing tooth occlusal contact area. Their edge attribute expression is: ,in, Indicates the first The and the first Interlocking edges between functional surface regions This indicates the static contact distance between the two regions. This represents the envelope surface of the dynamic occlusal sliding trajectory between two regions. Adjacent-surface associative edges are used to connect adjacent functional surface regions on a denture restoration; their edge attribute expression is... ,in, Indicates the first The and the first Adjacent face connections between functional face regions This indicates the length of the curve representing the common boundary between the two regions. This represents the curvature change feature vector of the common boundary curve between two regions. The dental prosthesis manufacturing terminal can automatically construct this topological relationship graph structure by calculating the spatial and anatomical relationships between the functional surface regions. This graph is used for process sequencing during subsequent toolpath generation and for judging the cascading effects of correcting a certain functional surface region on other related regions during AI reverse correction.

[0051] In one embodiment, a toolpath covering each functional surface region is generated based on a semantic design model based on functional semantic constraints, and semantic mapping information is attached to each tool point of each toolpath, including:

[0052] S31: Based on the unit cutting force coefficient of the denture restoration material, combined with the tool diameter, cutting depth and feed per tooth of the selected tool, calculate the theoretical value of the cutting force of each functional surface area under normal machining conditions, and construct the normal machining envelope range based on the theoretical value multiplied by the preset tolerance coefficient.

[0053] Indicatively, the dental prosthesis manufacturing terminal can obtain the unit cutting force coefficient corresponding to the prosthesis material. This unit cutting force coefficient can be pre-calibrated and stored through material cutting tests and used for calculating the theoretical value of the cutting force. The core formula for calculating the theoretical value of the cutting force is as follows: ,in, This represents the theoretical value of the main cutting force. This represents the unit cutting force coefficient of the dental prosthesis material, and its value is related to the physical properties of the material, such as hardness and toughness. This indicates the depth of cut, which is the depth to which the cutting tool penetrates the workpiece during the cutting process. This represents the feed per tooth, i.e., the distance the tool travels per tooth per revolution. The dental prosthesis manufacturing terminal can select different tool diameters, depths of cut, and feed per tooth based on the machining accuracy requirements of different functional areas, and then calculate the theoretical value of the main cutting force for each functional area. Optionally, to cover minor fluctuations during normal machining, the dental prosthesis manufacturing terminal can construct a normal machining envelope range based on the theoretical value of the main cutting force and a preset tolerance coefficient. The calculation formula is as follows: and ,in, This represents the upper bound of the normal processing envelope. This represents the lower bound of the normal processing envelope. This represents a preset tolerance coefficient, which the dental prosthesis manufacturing terminal can adjust based on processing experience and the characteristics of the denture material. Simultaneously, the dental prosthesis manufacturing terminal can derive the theoretical values ​​of the normal cutting force and the feed direction cutting force using cutting mechanics theory; the calculation formula is as follows: and ,in, This represents the theoretical value of the normal cutting force. This represents the theoretical value of the cutting force in the feed direction. and These represent the normal cutting force coefficient and the feed direction cutting force coefficient, respectively. Both are empirical coefficients based on experimental calibration, which are used to construct the normal machining envelope range of the cutting forces in the normal and feed directions.

[0054] S32: For each tool point on the tool path, calculate the nearest distance from the tool tip to the NURBS surface of each functional surface region in the semantic design model based on functional semantic constraints, and determine the functional surface region to which the tool point belongs by the nearest distance less than the preset spatial attribution threshold.

[0055] Optionally, after generating the toolpath, the dental prosthesis manufacturing terminal can calculate the shortest distance from the tool tip to the NURBS surface of all functional regions in the semantic design model based on functional semantic constraints for each tool point on the path, thereby determining the functional region to which the tool point belongs. The formula for calculating the shortest distance from the tool tip to the NURBS surface is as follows: ,in, Indicates the first The minimum nearest distance from the tool tip to the NURBS surface of each functional region at each tool position point. Indicates the first The spatial coordinate vector of the tool tip at each tool position point. Indicates the total number of functional surface regions. Indicates the first The NURBS surface expression for each functional region can be consistent with the general NURBS surface expression for denture restorations. This represents the Euclidean norm. For example, during the calculation process, the dental prosthesis manufacturing terminal can use a numerical iterative algorithm to solve for the nearest distance, with the iterative formula being: and ,in, and Indicates the first Parametric coordinates of the NURBS surface during the next iteration and Indicates the first The parameter coordinates at the next iteration and They represent the closest distances. For parameters and The first-order partial derivative, and and represent the second-order partial derivatives, respectively. The iteration termination condition is that the difference in parameter coordinates between two iterations is less than a preset iteration convergence threshold. Furthermore, after the iteration is complete, the shortest distance from the blade tip to the NURBS surface of each functional area is obtained. The dental prosthesis manufacturing terminal can set a preset spatial assignment threshold to determine the minimum shortest distance. Functional surface regions with distances less than the preset threshold are identified as the functional surface regions to which the cutting point belongs, ensuring a precise association between the cutting point and the functional surface region. If multiple functional surface regions exist whose nearest distance is less than the preset spatial assignment threshold, the dental prosthesis manufacturing terminal can combine the topological relationship diagram to select the functional surface region that best matches the cutting point's processing trajectory as the assigned functional surface region.

[0056] S33: Perform the projection point algorithm on the NURBS surface of each functional surface region to calculate the surface parameter coordinates corresponding to the projection points of the tool tip points on the NURBS surface of each tool position.

[0057] Specifically, after determining the functional surface region of each incision point, the dental prosthesis manufacturing terminal can perform a projection point algorithm on the NURBS surface of that functional surface region to solve for the projection point of the incision tip on the surface and obtain the corresponding surface parameter coordinates. Its projection condition formula is: ,in, Indicates the first The spatial coordinate vector of the tool tip at each tool position point. Indicates the first NURBS surface expression for each functional region. This indicates that the NURBS surface has the following parameters. The unit normal vector at that location, Let represent the vector dot product. This projection condition formula indicates that the line connecting the tip of the surface and the projection point is perpendicular to the surface normal vector at the projection point. The formula for calculating the unit normal vector of the surface is: ,in, and Representing NURBS surfaces respectively For parameters and The first-order partial derivative vector, Represents the cross product of vectors. The modulus of the vector is used to normalize the normal vector to a unit normal vector. The projection condition equations for dental prostheses can be solved using the Newton-Raphson iterative method, with the iterative formula being: ,in, Indicates the first The parameter coordinates of the next iteration Indicates the first The parameter coordinates of the next iteration Let the Jacobian matrix be a pair of projection condition functions. and The partial derivatives, Describes the inverse of the Jacobian matrix. This represents the projection condition function vector. The iteration terminates when the distance between the projection point and the blade tip is less than a preset projection accuracy threshold. After the iteration is complete, the obtained... This refers to the surface parameter coordinates corresponding to the projection point of the tool tip onto the NURBS surface of the functional area.

[0058] S34: Based on the region identifier code and surface parameter coordinates of the functional surface area, combined with the functional priority weight, normal error tolerance value, tangential error tolerance value and correction constraint function in the semantic label data, construct the semantic mapping information of the tool position.

[0059] Furthermore, the dental prosthesis manufacturing terminal can integrate relevant information about the functional surface region to construct semantic mapping information for each cutting point. This information clarifies the attribution relationship between the cutting point and the functional surface region and functional semantic constraints, providing support for processing anomaly detection and AI reverse correction. The semantic mapping information is represented as follows: ,in, Indicates the first Semantic mapping information of each knife point, Indicates the first The region identifier code of the functional surface area to which each tool point belongs is used to uniquely identify the functional surface area to which it belongs. Indicates the first The surface parameter coordinates corresponding to the projection points of the tool tip points on the NURBS surface of the functional area of ​​each tool position. This indicates the functional priority weight of the area to which the denture belongs; the larger the value, the more critical the area is to the function of the denture. and These represent the normal and tangential error tolerance values ​​for the functional surface region, respectively, used to define the permissible range of geometric errors in that region. This represents the set of modified constraint functions belonging to a functional surface region, containing all constraint terms corresponding to that region. The dental prosthesis manufacturing terminal can extract the region identifier code and semantic label data of the functional surface region from the semantic design model based on functional semantic constraints, and construct semantic mapping information by combining this with surface parameter coordinates.

[0060] Optionally, the dental prosthesis manufacturing terminal can associate and store semantic mapping information with geometric information such as the spatial coordinates of the tool position and the tool posture vector. The association formula is as follows: ,in, Indicates the first Complete information on each knife site, This represents the spatial coordinate vector of the tool tip at the tool position point. The tool attitude vector representing the tool position point. This represents the semantic mapping information of the tool position points. The dental prosthesis manufacturing terminal can write the complete information of the tool position points into the toolpath file, ensuring that the semantic mapping information can be called in real time during the machining process, and realizing the precise association between machining physical data and functional semantic constraints.

[0061] In one embodiment, the semantic mapping information of the machining physical data and the corresponding tool position points is spatiotemporally aligned and associated to generate a semantically enhanced machining physical data stream, including:

[0062] S41: Based on the spatial coordinates of the tool tip corresponding to each machining physical information record in the machining physical data, the semantic mapping index table based on the spatial grid hash structure is indexed to obtain the semantic mapping information corresponding to each machining physical information record; wherein, the semantic mapping index table based on the spatial grid hash structure uses the pose combination of each tool point as the index key and the semantic mapping information corresponding to each tool point as the index value; the pose combination includes the spatial coordinates of the tool tip and the tool posture vector.

[0063] Indicatively, the dental prosthesis manufacturing terminal can construct a semantic mapping index table based on a spatial grid hash structure to achieve rapid indexing and association between processing physical information records and semantic mapping information, improving the efficiency of spatiotemporal alignment. Specifically, the dental prosthesis manufacturing terminal can define the pose combination of the cutting tool points, expressed as follows: ,in, Indicates the first The pose combination of each knife point. This represents the spatial coordinate vector of the tool tip at the tool position point. This represents the tool attitude vector at the tool position point. The tool attitude vector can be represented using quaternions, and its expression is: ,in, Represent the real part of the quaternion. The imaginary part of the quaternion represents the spatial orientation of the tool. Furthermore, the dental prosthesis manufacturing terminal can construct a spatial grid hash structure, with the grid division formula being: ,in, Represents the spatial coordinates of the knife tip. The spatial grid number to which it belongs. Indicates the side length of the spatial grid. This represents a floor function, allowing the dental prosthesis fabrication terminal to adjust the grid side length based on the spatial distribution density of the cutting points. The value of is chosen to ensure index efficiency and association accuracy. Based on this spatial grid hash structure, the formula for constructing the semantic mapping index table is: ,in, This represents a semantic mapping index table based on a spatial grid hash structure. This represents a hash function, whose expression is: , This represents the XOR operation. The hash value representing the spatial grid number. The hash value representing the tool attitude vector. Indicates the first Semantic mapping information corresponding to each knife position.

[0064] Specifically, during the indexing process, the dental prosthesis manufacturing terminal can extract the spatial coordinates of the tool tip corresponding to each processing physical information record in the processing physical data. Combined with the tool attitude vector recorded during the machining process Construct the pose combination to be indexed Calculate its hash value using a hash function. The semantic mapping index table is searched to obtain the semantic mapping information corresponding to the processing physical information record. If hash collisions exist, the dental prosthesis manufacturing terminal can use a linked list method to resolve them. The collision resolution formula is as follows: ,in, This represents a linked list join function that stores conflicting semantic mapping information in the form of a linked list, ensuring the accuracy of the index.

[0065] S42: Associate each processing physical information record with its corresponding semantic mapping information to obtain semantic enhancement processing physical data records, and organize and sort the semantic enhancement processing physical data records according to the order of sampling time points to obtain semantic enhancement processing physical data stream.

[0066] Optionally, the dental prosthesis manufacturing terminal can associate each processing physical information record with its semantic mapping information obtained through indexing, constructing a semantically enhanced processing physical data record, thus achieving the organic integration of processing physical data and functional semantic constraint information. The processing physical information record is represented as follows: ,in, Indicates the first A record of physical information of the processing. This indicates the sampling timestamp of the record. This represents the cutting force signal vector at that sampling moment. This represents the vibration acceleration signal vector at that sampling moment. This represents the spindle power signal value at that sampling moment. The correlation formula for semantically enhanced machining physical data records is: ,in, Indicates the first A semantically enhanced physical data record, This represents the semantic mapping information corresponding to the physical information record of the processing. During the association process, the dental prosthesis manufacturing terminal can use a combination of timestamps and poses for dual verification to ensure the accuracy of the correspondence between the physical information record of the processing and the semantic mapping information. The verification formula is as follows: and ,in, This represents the difference between the sampling timestamp and the tool location execution timestamp. This indicates the runtime timestamp of the corresponding tool position. Indicates the time difference threshold. This represents the Euclidean distance between the pose combination to be associated and the tool position pose combination. This represents the pose distance threshold; only when both conditions are met is the association considered valid. After association is complete, the dental prosthesis manufacturing terminal can organize and sort all semantic enhancement processing physical data records according to the chronological order of sampling time. The objective function for sorting is... ,in, This represents the total number of physical data records processed for semantic enhancement. and They represent the first The and the first The sampling timestamps of each record are used to ensure the temporal continuity of the data stream through this sorting. After sorting, the dental prosthesis manufacturing terminal can organize the sorted semantic enhancement processing physical data records sequentially to form a semantic enhancement processing physical data stream, the expression of which is: The data stream contains processed physical data, semantic mapping information, and timestamp information.

[0067] In one embodiment, anomaly detection is performed on the semantically enhanced machining physical data stream. When the cutting force signal in any functional surface region deviates from the normal machining envelope range, the corresponding functional surface region is marked as a machining anomaly region, and the deviation vector of the machining physical data within the machining anomaly region is extracted, including:

[0068] S51: Compare each semantic enhancement processing physical data record carrying the same region identifier code in the semantic enhancement processing physical data stream with the upper and lower bounds of the normal processing envelope range point by point to obtain the comparison results.

[0069] In a schematic manner, the dental prosthesis manufacturing terminal can first group the semantic enhancement processing physical data stream, grouping semantic enhancement processing physical data records carrying the same region identifier into one group. Each group corresponds to a functional surface region, facilitating anomaly detection for different functional surface regions. For each semantic enhancement processing physical data record within each group, the dental prosthesis manufacturing terminal can extract the actual cutting force signal and calculate the actual cutting force composition. The calculation formula is as follows: ,in, Indicates the first The semantically enhanced machining physical data record contains the actual composite cutting force. This indicates the actual measured value of the main cutting force in the record. This indicates the actual measured value of the normal cutting force in the record. This represents the actual measured value of the cutting force in the feed direction recorded. Furthermore, the dental prosthesis manufacturing terminal can compare the actual composite cutting force with the upper and lower bounds of the normal machining envelope corresponding to the functional surface region point by point. The mathematical representation of the comparison result is as follows: ,in, Indicates the first The comparison results of each record This represents the upper bound of the normal processing envelope of the functional surface region. This represents the lower bound of the normal processing envelope of the functional surface region. When this occurs, it indicates that the actual resultant cutting force recorded deviates from the normal machining envelope range. This indicates that the actual combined cutting force of the recorded value is within the normal range. The dental prosthesis manufacturing terminal can statistically analyze the comparison results of all records within each group to obtain the overall comparative statistical results for that functional surface area. The statistical formula is as follows: ,in, This indicates the number of records that deviate from the normal range. This represents the set of physical data records for semantic enhancement processing corresponding to this functional area.

[0070] S52: When the comparison result shows that the actual cutting force in any semantically enhanced machining physical data record exceeds the normal machining envelope range, take the sampling time point corresponding to the semantically enhanced machining physical data record that exceeds the normal machining envelope range as the center, and extend the sampling points forward and backward along the tool path by a preset number of sampling points to obtain the abnormal sampling point interval.

[0071] Optionally, the dental prosthesis manufacturing terminal can judge the comparison results of each functional surface region. When it detects that any one of the comparison results for a certain functional surface region exists... When a record exists where the actual combined cutting force exceeds the normal machining envelope range, the process for expanding the abnormal sampling point interval is initiated. Specifically, the dental prosthesis manufacturing terminal can determine the sampling time point corresponding to the record that exceeds the range. And find the index position of that time point in the physical data stream of semantic enhancement processing. Furthermore, the dental prosthesis manufacturing terminal can be set with a preset expansion quantity. Extend along both the forward and backward directions of the toolpath The formula for the index range of the abnormal sampling point interval is: (The formula is missing from the original text.) ,in, This represents the set of indices for the interval of abnormal sampling points. Indicates the starting index of the interval of abnormal sampling points. This represents the terminating index of the abnormal sampling point interval. If the expanded starting index is less than 0, then 0 is used as the starting index; if the expanded terminating index is greater than the total number of physical data records for semantic enhancement processing... Then To terminate the index, its boundary correction formula is as follows: and ,in, This indicates the starting index of the corrected interval of outlier sampling points. This indicates the termination index of the corrected abnormal sampling point range. After the correction is completed, the dental prosthesis manufacturing terminal will index the range. All sampling points within the range are identified as abnormal sampling point intervals to ensure complete coverage of the abnormal area and its surrounding processing data.

[0072] S53: Aggregate all semantically enhanced processing physical data records within the abnormal sampling point interval into a processing abnormal region, and associate the spatial coordinates of the starting sampling point tool tip, the spatial coordinates of the ending sampling point tool tip, the functional surface region, and the statistical features of the semantically enhanced processing physical data of the processing abnormal region.

[0073] Specifically, the dental prosthesis manufacturing terminal can detect abnormal sampling point ranges. All semantic enhancement processing physical data records within the region are aggregated to form a processing anomaly region. The mathematical representation of the processing anomaly region is as follows: ,in, Indicates areas of abnormal processing. Indicates the first point within the interval of abnormal sampling points Each semantically enhanced physical data record. Furthermore, the dental prosthesis manufacturing terminal can extract correlation information of processing abnormal areas, including the spatial coordinates of the tool tip at the starting and ending sampling points, using the following extraction formula: , ,in, This represents the spatial coordinate vector of the tool tip point, indicating the starting sampling point of the machining anomaly region. Indicates that the index is The spatial coordinate vector of the blade tip corresponding to the sampling point. This represents the spatial coordinate vector of the tool tip point at the termination sampling point in the abnormal processing area. Indicates that the index is The sampling point corresponds to the spatial coordinate vector of the blade tip. Simultaneously, the dental prosthesis manufacturing terminal can determine the functional surface region corresponding to the processing abnormal area, and its correlation formula is... ,in, This represents the functional area corresponding to the area of ​​processing abnormality. This represents the functional area corresponding to the area identifier code shared by all records within the abnormal sampling point interval.

[0074] Optionally, the dental prosthesis manufacturing terminal can calculate the statistical characteristics of the semantically enhanced processing physical data, including the mean, variance, and peak value of the composite cutting force, calculated using the following formulas: , , ,in, This represents the average value of the resultant cutting force. This represents the variance of the resultant cutting force. This represents the peak value of the resultant cutting force. This indicates the number of sampling points within the interval of abnormal sampling points, i.e. The dental prosthesis manufacturing terminal can bind the above-mentioned related information with the processing abnormal area to form complete processing abnormal area information, the expression of which is: .

[0075] S54: Based on the semantically enhanced statistical features of machining physical data, extract the normal cutting force deviation and tangential cutting force deviation between the actual cutting force measurement value and the center value of the normal machining envelope of each sampling point in the abnormal machining area, and construct a deviation vector based on the combination of the normal cutting force deviation and tangential cutting force deviation of each sampling point.

[0076] Specifically, the dental prosthesis manufacturing terminal can extract the cutting force deviation at each sampling point based on the semantically enhanced processing physical data statistical features of the processing abnormal area, combined with the center value of the normal processing envelope. For example, the dental prosthesis manufacturing terminal can calculate the center value of the normal processing envelope using the following formula: ,in, This represents the center value of the normal processing envelope. This represents the upper bound of the normal processing envelope. This represents the lower bound of the normal machining envelope, whose center value is used as a standard reference value to characterize the cutting force under normal machining conditions. The dental prosthesis manufacturing terminal can extract the actual measured values ​​of the normal cutting force at each sampling point within the machining abnormality area. Actual measured value of tangential cutting force The normal cutting force deviation and the tangential cutting force deviation are calculated separately, and the calculation formula is as follows: and ,in, Indicates the first Normal cutting force deviation at each sampling point The center value of the normal machining envelope of the normal cutting force is represented, and its calculation method is the same as... Consistent Indicates the first Tangential cutting force deviation at each sampling point This represents the center value of the normal machining envelope of the tangential cutting force.

[0077] Optionally, to accurately reflect the degree and direction of processing abnormalities, the dental prosthesis manufacturing terminal can combine the normal and tangential cutting force deviations at each sampling point to construct a deviation vector, the expression of which is: ,in, Indicates the first The deviation vector of each sampling point This represents the transpose operation of a vector, and the deviation vector can simultaneously reflect the cutting force deviation in both the normal and tangential directions. Furthermore, the dental prosthesis manufacturing terminal can integrate the deviation vectors of all sampling points within the processing anomaly area to form a comprehensive deviation matrix for the processing anomaly area, the formula of which is: ,in The deviation matrix represents the machining abnormal area, which can comprehensively reflect the distribution characteristics of cutting force deviation within the machining abnormal area.

[0078] In one embodiment, the deviation vector of the processing anomaly region and the functional semantic constraints are jointly input into a pre-trained AI inverse correction model. The AI ​​inverse correction model solves within the geometric correction feasible region defined by the correction constraint function to obtain a geometric correction scheme for the processing anomaly region, including:

[0079] S61: Obtain the cumulative machining time and cumulative cutting length of the tool used in the current CNC machining process, evaluate the tool wear condition based on the cumulative machining time and cumulative cutting length, and generate the current tool wear condition parameters.

[0080] As an illustration, the dental prosthesis manufacturing terminal can collect real-time operating data of the tools used in the current CNC machining process, extracting two core parameters: cumulative machining time and cumulative cutting length, to assess tool wear. The formula for calculating the cumulative machining time is... ,in, This indicates the total machining time of the cutting tool. This indicates the number of machining tasks that the tool has completed. This indicates that the tool was in the first... The actual machining time in this machining task. The formula for calculating the cumulative cutting length is: ,in, This indicates the cumulative cutting length of the tool. Indicates the first The number of cutting path segments of the tool in the next machining task. Indicates the first In the next processing task The length of the cutting path segment. The dental prosthesis manufacturing terminal can generate tool wear state parameters based on cumulative machining time and cumulative cutting length through normalization processing. The calculation formula is as follows: ,in, The current tool wear state parameter is represented by a two-dimensional vector. This indicates the rated cumulative machining time for this type of cutting tool. This indicates the rated cumulative cutting length for this type of cutting tool. Dental prosthesis manufacturing terminals can adjust the rated parameters based on the tool's material, model, and machining conditions. and The value of is determined to ensure the accuracy of tool wear condition assessment. At the same time, the tool wear condition parameters can be compared with historical wear data to further optimize the assessment results.

[0081] S62: Convert the functional surface type identifier of the functional surface region corresponding to the processing abnormal region into a dense vector through the embedding layer, and extract the geometric feature vector from the control point coordinate matrix in the geometric representation data of the corresponding functional surface region through a multilayer perceptron.

[0082] Furthermore, the dental prosthesis manufacturing terminal can extract features from the functional surface regions corresponding to the processing abnormal areas. Specifically, the dental prosthesis manufacturing terminal can convert the functional surface type identifier into a dense vector through an embedding layer. The conversion formula for the embedding layer is as follows: ,in, This represents the transformed dense vector of function surface types. The function type identifier is represented in integer encoding. This represents the weight matrix of the embedding layer, with dimension 1. , Indicates the total number of functional facet types. The dimension representing a dense vector can be adjusted by the dental prosthesis manufacturing terminal according to the number of functional surface types. The value of .

[0083] Optionally, the dental prosthesis manufacturing terminal can extract the control point coordinate matrix from the geometric representation data of the functional surface region. After flattening it into a one-dimensional vector, it is input into a multilayer perceptron to extract geometric feature vectors. The forward propagation formula of the multilayer perceptron is: ,in, This represents the extracted geometric feature vector. This represents a one-dimensional vector after the control point coordinate matrix has been flattened. and These represent the weight matrices of the first and second hidden layers of the multilayer perceptron, respectively. and These represent the corresponding bias vectors. and These represent the activation functions of the first hidden layer and the output layer, respectively. The dental prosthesis manufacturing terminal can select activation functions such as ReLU and Sigmoid to ensure the effective extraction of geometric features.

[0084] S63: Extract the physical signal deviation time-series feature vector from the deviation vector through a one-dimensional convolutional layer, and transform the current tool wear state parameters into auxiliary feature vectors.

[0085] Specifically, the dental prosthesis manufacturing terminal can extract features from deviation vectors and tool wear parameters to obtain corresponding temporal and auxiliary features. For deviation vectors in areas of machining abnormalities... The dental prosthesis manufacturing terminal can input the data into a one-dimensional convolutional layer to extract the temporal feature vector of physical signal deviation. The calculation formula for the one-dimensional convolutional layer is as follows: ,in, This represents the extracted physical signal deviation time-series feature vector. This represents the kernel weight matrix of a one-dimensional convolutional layer, with dimensions of . , Indicates the length of the convolution kernel. This represents the number of channels in the input bias vector. This indicates the number of channels in the output time-series feature vector. Indicates the convolution stride. This indicates the filling method; the dental prosthesis manufacturing terminal can adjust the convolution kernel length and stride values ​​based on the length of the deviation vector. For the current tool wear state parameters... The dental prosthesis manufacturing terminal can transform it into an auxiliary feature vector through a fully connected layer. The transformation formula is as follows: ,in, Represents the auxiliary feature vector. This represents the weight matrix of the fully connected layer. This represents the bias vector of the fully connected layer. The dimension of the auxiliary feature vector is consistent with the dimensions of the geometric feature vector and the temporal feature vector.

[0086] S64: The dense vector, geometric feature vector, physical signal deviation time-series feature vector, and auxiliary feature vector are fused in the encoder intermediate layer of the AI ​​inverse correction model to obtain a semantic latent space vector that represents the mapping relationship between the abnormal processing state and the functional surface attributes. The decoder then reconstructs the preliminary correction amount of the coordinates of each control point of the NURBS surface of the functional surface region from the semantic latent space vector.

[0087] Optionally, the dental prosthesis manufacturing terminal can perform feature fusion on the extracted feature vectors in the intermediate layer of the encoder of the AI ​​inverse correction model to construct a semantic latent space vector. The calculation formula for feature fusion is as follows: ,in, Represents a semantic latent space vector. This represents the weight matrix of the feature fusion layer. This represents the bias vector of the feature fusion layer. This represents the transpose operation of a vector. This represents the joint feature vector formed by concatenating four types of feature vectors. Semantic latent space vector. It can accurately characterize the mapping relationship between abnormal processing states and functional surface attributes, providing support for the reconstruction of subsequent preliminary correction amounts. Furthermore, the dental prosthesis manufacturing terminal can input semantic latent space vectors into the decoder of the AI ​​inverse correction model. The decoder reconstructs the preliminary correction amounts of the coordinates of each control point on the NURBS surface of the functional surface region. The decoder's output formula is... ,in, The matrix representing the initial correction values ​​for the coordinates of each control point, with dimensions equal to those of the control point coordinate matrix. Consistent This represents the weight matrix of the decoder. This represents the bias vector of the decoder. The decoder can be constructed using deconvolutional layers or multilayer perceptrons to ensure that the initial correction amount can reflect the geometric deviations corresponding to processing anomalies.

[0088] S65: Construct a quadratic optimization problem using the correction constraint function of the functional surface region corresponding to the abnormal processing area as the constraint condition, and substitute the initial correction amount into the quadratic optimization problem to perform constraint projection operation to obtain the geometric correction scheme; The constraint projection operation takes minimizing the Euclidean distance between the adjusted control point coordinate correction amount and the initial correction amount as the objective function, and the correction constraint function as the boundary of the feasible region. The final control point coordinate correction amount is obtained by solving the quadratic optimization problem, and the geometric correction scheme is constructed based on the final control point coordinate correction amount.

[0089] For example, the dental prosthesis manufacturing terminal can construct a quadratic optimization problem using the correction constraint function of the functional surface region corresponding to the processing abnormal area as the constraint condition. The initial correction amount is adjusted through constraint projection operations to obtain the final correction amount that meets the constraint requirements. Specifically, the objective function of the quadratic optimization problem is... ,in, This represents the final control point coordinate correction matrix. This represents the initial correction matrix. Let represent the square of the Euclidean distance. This objective function aims to minimize the deviation between the final correction and the initial correction, reducing unnecessary adjustments. The constraints of the quadratic optimization problem are all the correction constraint functions for the functional surface region corresponding to the processing anomaly region, mathematically expressed as: ,in, This represents the set of correction constraint functions for the functional surface region. This represents the control point coordinate matrix before correction. This represents the corrected control point coordinate matrix, and the constraints ensure that the corrected NURBS surface meets the functional semantic constraints. The interior point method can be used to solve this quadratic optimization problem in the dental prosthesis manufacturing process, and its iterative formula is: ,in, Indicates the first The final correction matrix of the next iteration Indicates the first The final correction matrix of the next iteration The Hessian matrix represents the objective function. This represents the gradient vector of the objective function. The iteration terminates when the magnitude of the gradient vector is less than a preset optimization threshold. After solving, the final control point coordinate correction matrix is ​​obtained. The dental prosthesis manufacturing terminal can construct a geometric correction scheme based on this final correction amount, the expression of which is: ,in, This represents the corrected control point coordinate matrix. This represents the corrected functional surface region NURBS surface. This geometric correction scheme can accurately correct geometric deviations in areas of abnormal processing while meeting functional semantic constraints.

[0090] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0091] In one embodiment, a computer device is provided, including a memory and a processor, the memory storing a computer program, the processor executing the computer program to implement the steps in the above method embodiments.

[0092] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps in the above method embodiments.

[0093] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The components described as separate parts may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this disclosure according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0094] The above-described embodiments are merely illustrative of several implementation methods of the embodiments of this application, and their descriptions are relatively specific and detailed. However, they should not be construed as limiting the scope of the patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the embodiments of this application, and these modifications and improvements all fall within the protection scope of the embodiments of this application.

Claims

1. A closed-loop management method for the entire process of AI-assisted design and CNC machining of dentures, characterized in that, The method includes: S1: Acquire digital impression data of the patient's oral cavity, and perform tooth segmentation and occlusal relationship identification based on the digital impression data to generate a three-dimensional model of the prosthesis; S2: Based on prior anatomical knowledge, multiple functional surface regions are divided on the three-dimensional model of the prosthesis, and a semantic design model based on functional semantic constraints is constructed for the functional surface regions; wherein, the functional semantic constraints include the geometric error tolerance range corresponding to each functional surface region and the correction constraint function used to limit the feasible domain of geometric correction; S3: Generate toolpaths covering each functional surface region according to the semantic design model based on functional semantic constraints, and attach semantic mapping information to each tool point of each toolpath; the semantic mapping information is used to characterize the attribution association between the tool point and the corresponding functional surface region and the functional semantic constraint respectively; S4: Acquire the machining physical data collected in real time during CNC machining, and perform spatiotemporal alignment and association between the machining physical data and the semantic mapping information of the corresponding tool position point to generate a semantically enhanced machining physical data stream; wherein, the machining physical data includes cutting force signals; S5: Perform anomaly detection on the semantically enhanced machining physical data stream. When the cutting force signal in any of the functional surface regions deviates from the normal machining envelope range, mark the corresponding functional surface region as a machining anomaly region and extract the deviation vector of the machining physical data in the machining anomaly region. S6: Input the deviation vector and the functional semantic constraint of the processing anomaly region into the pre-trained AI inverse correction model. The AI ​​inverse correction model solves within the geometric correction feasible region defined by the correction constraint function to obtain the geometric correction scheme of the processing anomaly region. S7: Apply the geometric correction scheme to the corresponding functional surface region in the semantic design model based on functional semantic constraints to obtain the corrected semantic design model based on functional semantic constraints, and cyclically execute S3 to S6 based on the corrected semantic design model based on functional semantic constraints until all deviation vectors in the processing abnormal regions fall within the geometric error tolerance range of the corresponding functional surface region.

2. The method according to claim 1, characterized in that, The construction of a semantic design model based on functional semantic constraints for each of the aforementioned functional surface regions includes: S21: Based on the three-dimensional model of the prosthesis, the NURBS surface representation of the prosthesis is obtained by using the boundary curves of each functional surface region as trimming curves to extract sub-surface patches of each functional surface region, and the control point coordinate matrix, node vector sequence, weight sequence and surface order of each sub-surface patch are saved as geometric representation data of the functional surface region. S22: Create semantic label data for each of the functional surface regions; the semantic label data includes a functional surface type identifier, functional priority weight, normal error tolerance value, tangential error tolerance value, and the correction constraint function; wherein, if the functional surface type identifier is an occlusal contact surface type, the correction constraint function of the functional surface region includes a first constraint term and a second constraint term; if the functional surface type identifier is an edge-closed surface type, the correction constraint function of the functional surface region includes a third constraint term and a fourth constraint term; the first constraint term is used to constrain the corrected sub-surface not to penetrate the opposing tooth surface within the contact area; the second constraint term is used to constrain the average curvature change of the corrected sub-surface not to exceed a preset curvature change threshold within the central contact sub-region; the third constraint term is used to constrain the corrected sub-surface not to deviate from the abutment tooth preparation termination line surface by more than a preset distance threshold within the edge-closed region; the fourth constraint term is used to constrain the local thickness of the corrected sub-surface not to be less than a preset minimum thickness value within the edge-closed region. S23: Combine the geometric representation data of each functional surface region with the semantic label data into a dual-channel data structure; S24: Aggregate the dual-channel data structures of all functional surface regions into the semantic design model based on functional semantic constraints, and establish a topological association graph structure in the semantic design model based on functional semantic constraints to represent the interlocking association relationship and the adjacent surface association relationship between each functional surface region.

3. The method according to claim 2, characterized in that, The step of generating toolpaths covering each functional surface region based on the semantic design model based on functional semantic constraints, and attaching semantic mapping information to each tool point of each toolpath, includes: S31: Based on the unit cutting force coefficient of the denture restoration material, combined with the tool diameter, cutting depth and feed per tooth of the selected tool, calculate the theoretical value of the cutting force of each functional surface area under normal machining conditions, and construct the normal machining envelope range based on the theoretical value multiplied by the preset tolerance coefficient. S32: For each tool point on the tool path, calculate the nearest distance from the tool tip to the NURBS surface of each functional surface region in the semantic design model based on functional semantic constraints, and determine the functional surface region to which the tool point belongs if the nearest distance is less than a preset spatial attribution threshold. S33: Perform a projection point algorithm on the NURBS surface of each of the said functional surface regions to calculate the surface parameter coordinates corresponding to the projection points of the tool tip points of each of the said tool points on the NURBS surface; S34: Based on the region identifier code of the functional surface area, the surface parameter coordinates, and combined with the functional priority weight, normal error tolerance value, tangential error tolerance value and correction constraint function in the semantic tag data, construct the semantic mapping information of the tool position.

4. The method according to claim 3, characterized in that, The step of spatiotemporally aligning and associating the machining physical data with the corresponding semantic mapping information of the tool position to generate a semantically enhanced machining physical data stream includes: S41: Based on the spatial coordinates of the tool tip corresponding to each machining physical information record in the machining physical data, indexing is performed in the semantic mapping index table based on the spatial grid hash structure to obtain the semantic mapping information corresponding to each machining physical information record; wherein, the semantic mapping index table based on the spatial grid hash structure uses the pose combination of each tool position as the index key and the semantic mapping information corresponding to each tool position as the index value; the pose combination includes the spatial coordinates of the tool tip and the tool posture vector; S42: Associate each of the processing physical information records with the corresponding semantic mapping information to obtain semantic enhancement processing physical data records, and organize and sort the semantic enhancement processing physical data records according to the order of sampling time points to obtain the semantic enhancement processing physical data stream.

5. The method according to claim 4, characterized in that, The anomaly detection of the semantically enhanced machining physical data stream, when the cutting force signal in any of the functional surface regions deviates from the normal machining envelope range, marks the corresponding functional surface region as a machining anomaly region, and extracts the deviation vector of the machining physical data within the machining anomaly region, including: S51: Compare each of the semantic enhancement processing physical data records carrying the same region identifier code in the semantic enhancement processing physical data stream with the upper and lower bounds of the normal processing envelope range point by point to obtain the comparison result; S52: When the comparison result is that the actual cutting force in any of the semantically enhanced machining physical data records exceeds the range of the normal machining envelope, take the sampling time point corresponding to the semantically enhanced machining physical data record that exceeds the range of the normal machining envelope as the center, and extend the sampling points forward and backward along the direction of the tool path by a preset number of sampling points to obtain the abnormal sampling point interval. S53: Aggregate all the semantic enhancement processing physical data records within the abnormal sampling point interval into the processing abnormal region, and associate the starting sampling point tool tip spatial coordinates, the ending sampling point tool tip spatial coordinates, the functional surface region, and the semantic enhancement processing physical data statistical feature quantity of the processing abnormal region. S54: Based on the semantically enhanced machining physical data statistical features, extract the normal cutting force deviation and tangential cutting force deviation between the actual cutting force measurement value and the center value of the normal machining envelope of each sampling point in the abnormal machining area, and construct the deviation vector based on the combination of the normal cutting force deviation and tangential cutting force deviation of each sampling point.

6. The method according to claim 5, characterized in that, The deviation vector and the functional semantic constraints of the processing anomaly region are jointly input into a pre-trained AI inverse correction model. The AI ​​inverse correction model solves within the geometric correction feasible region defined by the correction constraint function to obtain a geometric correction scheme for the processing anomaly region, including: S61: Obtain the cumulative machining time and cumulative cutting length of the tool used in the current CNC machining process, evaluate the tool wear state based on the cumulative machining time and cumulative cutting length, and generate the current tool wear state parameters; S62: Convert the functional surface type identifier of the functional surface region corresponding to the processing abnormal region into a dense vector through an embedding layer, and extract the geometric feature vector of the control point coordinate matrix in the geometric representation data of the corresponding functional surface region through a multilayer perceptron. S63: Extract the physical signal deviation time-series feature vector from the deviation vector through a one-dimensional convolutional layer, and convert the current tool wear state parameter into an auxiliary feature vector; S64: The dense vector, the geometric feature vector, the physical signal deviation time-series feature vector, and the auxiliary feature vector are fused in the encoder intermediate layer of the AI ​​inverse correction model to obtain a semantic latent space vector that characterizes the mapping relationship between the abnormal processing state and the functional surface attributes. The decoder then reconstructs the preliminary correction amount of the coordinates of each control point of the NURBS surface of the functional surface region from the semantic latent space vector. S65: Construct a quadratic optimization problem using the correction constraint function of the functional surface region corresponding to the abnormal processing region as the constraint condition, and substitute the initial correction amount into the quadratic optimization problem to perform a constraint projection operation to obtain the geometric correction scheme; the constraint projection operation takes minimizing the Euclidean distance between the adjusted control point coordinate correction amount and the initial correction amount as the objective function, and the correction constraint function as the feasible region boundary, and obtains the final control point coordinate correction amount by solving the quadratic optimization problem, and constructs the geometric correction scheme based on the final control point coordinate correction amount.

7. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the method of any one of claims 1 to 6.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method of any one of claims 1 to 6.

Images