Metal 3D printing quality prediction method and system

Abstract

The invention relates to the technical field of 3D printing quality prediction, and particularly discloses a metal 3D printing quality prediction method and system.The method comprises the steps that real-time state information of a 3D printer and state data of a current printing piece are collected; calculating the thermal stress of each layer of printed piece in the printing process; combining a convolutional neural network and a time sequence neural network with regional clustering to generate a spatial geometric accuracy grade map; in the prior art, quality analysis is carried out mostly through single temperature monitoring or post-processing detection, and especially under the printing conditions that the structure is complex and interlayer stress coupling is remarkable, precision prediction preposition and real-time space error analysis are difficult to achieve. Due to the fact that thermal stress driven two-dimensional error modeling, three-dimensional error distribution prediction based on deep learning and clustering type space grade division are introduced, process-level prediction and structure-level expression of the printing piece quality are achieved, and prediction accuracy and space expression capacity are improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of 3D printing quality prediction, and in particular relates to a metal 3D printing quality prediction method and system. Background Art

[0002] Currently, metal 3D printing, as a key branch of additive manufacturing technology, has been widely used in aerospace, medical devices, high-end equipment, and other manufacturing fields with extremely high requirements for structural performance and precision. During the printing process, geometric accuracy is a key indicator of printed part quality. It not only determines whether the printed part meets the designed dimensions and structural tolerances, but also directly determines the feasibility of subsequent assembly and the reliability of functional implementation. Especially in precision components with complex internal cavities, free-form surfaces, and multi-layer detailed structures, the accumulation of slight spatial errors can lead to assembly failure or performance abnormalities. However, existing metal 3D printing quality evaluation methods mostly rely on dimensional measurement or mechanical testing in the post-processing stage, which cannot predict accuracy trends and perceive regional errors during the printing process. For example, some methods only make rough judgments based on single-point temperature monitoring or single-layer stress analysis, failing to fully consider important factors affecting geometric accuracy such as multi-layer thermal coupling, material thermal deformation behavior, and printing path disturbances. This makes it difficult to meet the requirements for predictable and interpretable spatial accuracy throughout the entire process in practical applications. Therefore, there is an urgent need for a quality prediction method that can integrate the thermal stress evolution law and printing status data without relying on additional sensors or interrupting the printing process, and can predict and grade the geometric errors of the printed parts at various positions in space in advance, so as to improve the geometric accuracy assurance capability and overall manufacturing reliability of metal 3D printing. Summary of the Invention

[0003] In response to the above-mentioned technical deficiencies, the purpose of the present invention is to propose a metal 3D printing quality prediction method, which aims to solve the technical problem that the existing technology mostly uses single temperature monitoring or post-processing detection for quality analysis, especially under printing conditions with complex structures and significant interlayer stress coupling, and it is difficult to achieve accurate prediction pre-processing and real-time spatial error analysis.

[0004] In order to solve the above technical problems, the present invention adopts the following technical solutions: The present invention provides a metal 3D printing quality prediction method,

[0005] The metal 3D printing quality prediction method includes:

[0006] Step S10: collecting real-time status information of the 3D printer and current print status data, and calculating the current print thermal stress based on the real-time status information and the current print status data;

[0007] Step S20: obtaining a target design size of a target printed part, and constructing a two-dimensional error map according to the target design size and the thermal stress of the current printed part;

[0008] Step S30: The two-dimensional error map, real-time status information, and current print status data are normalized to obtain a fused input vector, a preset error distribution prediction model is pre-set, and the error distribution prediction model is fused and trained using a convolutional neural network and a time series network. The fused input vector is input into the trained error distribution prediction model to obtain a three-dimensional error distribution.

[0009] Step S40: performing regional clustering and error level classification based on the three-dimensional error distribution to generate a spatial geometric accuracy level map of the target print;

[0010] Step S50: generating a print quality prediction report based on the spatial geometric accuracy level map and the target design dimensions of the target print part.

[0011] Preferably, in step S10, the step of collecting the real-time status information of the 3D printer and the current print status data, and calculating the thermal stress of the current print according to the real-time status information and the current print status data specifically includes:

[0012] Collect the real-time status information of the 3D printer and the current print status data, and obtain the melting zone temperature T at the time t position (x, y) from the current print status data melt (t, x, y), the thermal expansion coefficient α of the metal material and the elastic modulus E of the metal material, and obtain the printer cavity ambient temperature at the time t position (x, y) from the real-time status information;

[0013] According to the melting zone temperature T melt (t, x, y) and the ambient temperature of the printer cavity are used to calculate the relative temperature difference;

[0014] The global adjustment factor of the thermal stress of the current printed part is introduced, and the thermal stress of the current printed part is calculated based on the relative difference temperature, the global adjustment factor of the thermal stress of the current printed part, the thermal expansion coefficient α of the metal material, and the elastic modulus E of the metal material.

[0015] Preferably, in step S10, the real-time status information includes printer temperature, printer laser power and printer scanning speed; the current print status data includes metal material type, metal material thermal expansion coefficient, metal material elastic modulus, melting zone temperature and current print thermal stress.

[0016] Preferably, in step S30, the two-dimensional error map, real-time status information, and current print status data are normalized to obtain a fused input vector, an error distribution prediction model is preset, and the error distribution prediction model is fused and trained using a convolutional neural network and a time series network. The fused input vector is input into the trained error distribution prediction model to obtain a three-dimensional error distribution, which specifically includes:

[0017] Step S301: normalizing the two-dimensional error map, the real-time status information, and the current print status data to construct a fusion input vector in a unified numerical space;

[0018] Step S302: using a convolutional neural network to extract local stress variation features and spatial correlation pattern features between geometric deformations from the fused input vector;

[0019] Step S303: extracting temporal correlation pattern features during the printing process from the fused input vector using a time series network;

[0020] Step S304: fusing the spatial correlation pattern features between the local stress variation features and the geometric deformation extracted by the convolutional neural network and the temporal correlation pattern features extracted by the time series network to obtain a feature fusion vector;

[0021] Step S305: Preset the error distribution prediction model, obtain historical error distribution annotation data and historical feature fusion vector data, and use the historical error distribution annotation data and historical feature fusion vector data as input of the error distribution prediction model to pre-train the error distribution prediction model. Finally, pass the feature fusion vector through the inter-layer fusion layer and the fully connected layer of the pre-trained error distribution prediction model to output the three-dimensional error distribution.

[0022] Preferably, in step S30, the inter-layer fusion layer is used to achieve temporal fusion from the layered two-dimensional error map to the overall three-dimensional error distribution, wherein the formula of the inter-layer fusion layer is expressed as:

[0023]

[0024] in, is the predicted geometric error of the k-th layer position (x, y), where z k is the height coordinate value of the k-th layer position (x, y); ΔL (k) (x,y) is the two-dimensional error map of the k-th layer position (x,y), h k-1 is the hidden state passed through the k-1th layer of the error distribution prediction model, s k is the process parameter feature vector corresponding to the kth layer, f fusion (·) is the fusion function of the error distribution prediction model.

[0025] Preferably, in step S40, the step of performing regional clustering and error level division according to the three-dimensional error distribution to generate a spatial geometric accuracy level map of the target print comprises:

[0026] Step S401: performing voxel segmentation on the three-dimensional error distribution to obtain a plurality of spatial sub-regions, where the spatial sub-regions are used to represent the geometric deviation of each point of the target print in the three-dimensional space;

[0027] Step S402: extracting local error statistical features of each spatial sub-region using a mathematical statistical method to form an error feature vector;

[0028] Step S403: performing unsupervised cluster analysis on the multiple spatial sub-regions based on the error feature vector to obtain multiple cluster sub-regions;

[0029] Step S404: obtaining a preset error level classification standard, performing level mapping on multiple cluster sub-regions according to the preset error level classification standard, and generating a spatial geometric accuracy level map of the target print.

[0030] Preferably, in step S20, the formula for the two-dimensional error map is expressed as:

[0031]

[0032] Where σ(t,x,y) is the thermal stress of the current printed part at position (x,y) at time t, ΔL(x,y) is a two-dimensional error map representing the linear geometric deformation at position (x,y), L0(x,y) is the target design dimension at position (x,y), E is the elastic modulus of the metal material, and δ is the global adjustment factor for linear geometric deformation.

[0033] The present invention also provides a metal 3D printing quality prediction system comprising:

[0034] A state stress calculation module is used to collect real-time state information of the 3D printer and current print status data, and calculate the current print thermal stress based on the real-time state information and current print status data;

[0035] An error mapping module is used to obtain the target design size of the target printed part and construct a two-dimensional error map based on the target design size and the thermal stress of the current printed part;

[0036] The error prediction modeling module is used to obtain a fused input vector after normalizing the two-dimensional error map, real-time status information, and current print status data, preset an error distribution prediction model, and use a convolutional neural network and a time series network to train the error distribution prediction model. The fused input vector is input into the trained error distribution prediction model to obtain a three-dimensional error distribution;

[0037] A grade determination module is used to perform regional clustering and error grade classification based on the three-dimensional error distribution, and generate a spatial geometric accuracy grade map of the target print;

[0038] The quality prediction output module is used to generate a print quality prediction report based on the spatial geometric accuracy level map and the target design size of the target print part.

[0039] The present invention also provides a computer program product, including a metal 3D printing quality prediction program, which implements the metal 3D printing quality prediction method when executed by a processor.

[0040] The beneficial effect of the present invention is that compared with the existing technology that mostly performs quality analysis through single temperature monitoring or post-processing detection, especially under printing conditions with complex structures and significant interlayer stress coupling, it is difficult to achieve the technical problem of precision prediction pre-processing and real-time spatial error analysis. Since the present application introduces thermal stress-driven two-dimensional error modeling, deep learning-based three-dimensional error distribution prediction and clustered spatial hierarchical division, it realizes process-level prediction and structural-level expression of print quality, thereby improving prediction accuracy and spatial expression capabilities. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0042] Figure 1 This is a flow chart of a first embodiment of a metal 3D printing quality prediction method according to the present invention.

[0043] Figure 2 Schematic diagram of equipment for a metal 3D printing quality prediction method of the present invention. DETAILED DESCRIPTION

[0044] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0045] Example 1: Figure 1 FIG. 1 is a flow chart of the first embodiment of the metal 3D printing quality prediction method of the present invention, which provides the first embodiment of the metal 3D printing quality prediction method of the present invention.

[0046] In a first embodiment, the metal 3D printing quality prediction method includes:

[0047] Step S10: collecting real-time status information of the 3D printer and current print status data, and calculating the current print thermal stress based on the real-time status information and the current print status data;

[0048] It should be noted that in step S10, the calculation formula for the thermal stress of the current printed part is:

[0049] σ(t,x,y)=γ·E·α·(T melt (t,x,y)-T chamber (t,x,y))

[0050] Where γ is the global adjustment factor of the thermal stress of the current printed part, σ(t,x,y) is the thermal stress of the current printed part at the time t position (x,y); E is the elastic modulus of the metal material; α is the thermal expansion coefficient of the metal material; T melt (t, x, y) is the temperature of the melting zone at position (x, y) at time t, which is determined based on the current print status data; T chamber (t, x, y) is the ambient temperature of the printing chamber at position (x, y) at time t, which is determined based on the real-time status information.

[0051] It can be understood that this formula effectively captures the trend of local thermal stress changes by introducing the temperature difference terms of the two variables of time and space, while keeping the physical properties of the material unchanged, and can make advance estimates of potential stress concentration phenomena in different areas.

[0052] It should be understood that compared with traditional methods that only consider average heat input or static temperature difference estimation, this method introduces a dynamic difference term between the printer ambient temperature and the melting temperature in thermal stress prediction, which can more realistically reflect the stress response behavior of the material under non-uniform temperature fields. It is particularly suitable for metal 3D printing scenarios with complex path scanning and non-constant temperature cavity conditions.

[0053] For example, when printing a titanium alloy structure with a significant wall thickness change, by setting the material parameters E = 110GPa, α = 8.5×10 -6 / ℃, when the local melting zone temperature is 1550℃, the cavity temperature is 100℃, and the global adjustment factor is set to 0.9, the predicted thermal stress is: σ=0.9·110×10 9 8.5×10 -6 (1550-100)≈135 MPa. This predicted value is consistent with the trend of residual stress field measured by subsequent thermal deformation experiments, verifying the effectiveness of the constructed thermal stress model.

[0054] Step S20: obtaining a target design size of a target printed part, and constructing a two-dimensional error map according to the target design size and the thermal stress of the current printed part;

[0055] It should be noted that, in step S20, the formula for the two-dimensional error map is expressed as:

[0056]

[0057] Where σ(t,x,y) is the thermal stress of the current printed part at position (x,y) at time t, ΔL(x,y) is a two-dimensional error map representing the linear geometric deformation at position (x,y), L0(x,y) is the target design dimension at position (x,y), E is the elastic modulus of the metal material, and δ is the global adjustment factor for linear geometric deformation.

[0058] It's understandable that the error map construction method described above is based on the linear strain response relationship in elastic mechanics. Assuming the material is in the elastic deformation stage, the stress field drives the estimation of small changes in the target dimension, thus forming a regional error field. This error map spatially retains its mapping relationship with the CAD design model, facilitating subsequent spatial splicing and predictive modeling.

[0059] It should be understood that compared to traditional estimation methods based on scan path superposition or statistical average error, this method, which starts from the microscopic field response of thermal stress and establishes a deformation mapping relationship based on target size, can achieve spatially continuous error prediction and has higher physical interpretation power and structural adaptability. This is particularly true for complex prints with porous structures and variable thickness layer designs, significantly improving the ability to predict boundary deformation and local deviations.

[0060] For example, the predicted thermal stress value at a certain point is σ = 120 MPa, the material elastic modulus is E = 200 GPa, the adjustment factor δ = 0.95, and L0(x, y) = 40 mm. Substituting the formula into the error at this point is:

[0061]

[0062] The predicted error value is basically consistent with the actual error (23.1μm) detected by the three-coordinate measuring machine after printing, verifying the effectiveness of the error map construction method at micron-level accuracy.

[0063] Step S30: The two-dimensional error map, real-time status information, and current print status data are normalized to obtain a fused input vector, a preset error distribution prediction model is pre-set, and the error distribution prediction model is fused and trained using a convolutional neural network and a time series network. The fused input vector is input into the trained error distribution prediction model to obtain a three-dimensional error distribution.

[0064] It should be noted that in step S30, the two-dimensional error map, real-time status information, and current print status data are normalized to obtain a fused input vector, an error distribution prediction model is preset, and the error distribution prediction model is fused and trained using a convolutional neural network and a time series network. The fused input vector is input into the trained error distribution prediction model to obtain a three-dimensional error distribution. The steps specifically include:

[0065] Step S301: normalizing the two-dimensional error map, the real-time status information, and the current print status data to construct a fusion input vector in a unified numerical space;

[0066] Step S302: using a convolutional neural network to extract local stress variation features and spatial correlation pattern features between geometric deformations from the fused input vector;

[0067] Step S303: extracting temporal correlation pattern features during the printing process from the fused input vector using a time series network;

[0068] Step S304: fusing the spatial correlation pattern features between the local stress variation features and the geometric deformation extracted by the convolutional neural network and the temporal correlation pattern features extracted by the time series network to obtain a feature fusion vector;

[0069] Step S305: Preset the error distribution prediction model, obtain historical error distribution annotation data and historical feature fusion vector data, and use the historical error distribution annotation data and historical feature fusion vector data as input of the error distribution prediction model to pre-train the error distribution prediction model. Finally, pass the feature fusion vector through the inter-layer fusion layer and the fully connected layer of the pre-trained error distribution prediction model to output the three-dimensional error distribution.

[0070] This modeling approach, through deep learning's joint spatial-temporal feature extraction mechanism, addresses the issues of spatial and process decoupling and local prediction discontinuity found in traditional models. The fused three-dimensional error distribution not only captures the local thermal stress response of the current layer but also reflects the influence of the thermal history of layers above and below.

[0071] It should be understood that compared with the existing technology that only relies on surface point error regression or layer-by-layer error stacking, the present invention realizes continuous prediction of three-dimensional errors by introducing the inter-layer fusion mechanism of convolutional neural networks and time series networks, and supports end-to-end data-driven training. It automatically obtains the error evolution law without manually setting rules, and significantly improves adaptability and prediction accuracy.

[0072] For example, in the error prediction task of a printed part with a highly complex topological structure, after using 800 sets of historical printing samples to train the model, the new two-dimensional error map (average layer stress 90MPa) and process data (laser power, scanning speed, material type) were input into the model as fusion vectors. The output three-dimensional error distribution had an average error of only 21μm compared with the actual measurement value on the test set. Compared with the traditional error stacking method (error of about 41μm), the prediction error was reduced by about 48.8%, the spatial coherence was stronger, and the accuracy distribution was closer to the actual thermal stress deformation trend.

[0073] Step S40: performing regional clustering and error level classification based on the three-dimensional error distribution to generate a spatial geometric accuracy level map of the target print;

[0074] It should be noted that, in step S401 , the three-dimensional error distribution is voxelized and divided to obtain a plurality of spatial sub-regions, where the spatial sub-regions are used to represent the geometric deviation of each point of the target print in the three-dimensional space;

[0075] Step S402: extracting local error statistical features of each spatial sub-region using a mathematical statistical method to form an error feature vector;

[0076] Step S403: performing unsupervised cluster analysis on the multiple spatial sub-regions based on the error feature vector to obtain multiple cluster sub-regions;

[0077] Step S404: obtaining a preset error level classification standard, performing level mapping on multiple cluster sub-regions according to the preset error level classification standard, and generating a spatial geometric accuracy level map of the target print.

[0078] It should be noted that the voxelization of the three-dimensional error distribution is specifically to evenly divide it into several equal-sized cubic sub-blocks along the three spatial dimensions. Each sub-block contains the local error values ​​of several points, which serve as the spatial minimum unit for subsequent statistical analysis to ensure the spatial resolution consistency and the locality of the error expression; the mathematical statistical method is used to extract the local error statistical characteristics of each spatial sub-region, specifically to extract the mathematical features of average error, maximum error, standard deviation and local gradient in each sub-block; the unsupervised cluster analysis specifically includes K-means cluster analysis and DBSCAN cluster analysis; the preset error level classification standard defines the error indicator threshold range corresponding to each level, such as the classification standards for average error, maximum error, error stability, etc., and then the level mapping of multiple cluster sub-regions specifically includes comparing and judging the error characteristics, and marking them with corresponding level labels based on whether their characteristic values ​​fall into the range corresponding to a certain level standard. The level label is used to indicate the geometric accuracy risk level of the spatial sub-region, such as "high precision area", "general area", "warning area", etc. After all cluster sub-regions have completed the level mapping, the accuracy level map of the entire print in three-dimensional space can be constructed to achieve visualization and regional expression of printing errors.

[0079] This modeling approach, through deep learning's joint spatial-temporal feature extraction mechanism, addresses the issues of spatial and process decoupling and local prediction discontinuity found in traditional models. The fused three-dimensional error distribution not only reflects the local thermal stress response of the current layer but also the influence of the thermal history of layers above and below it. This prediction is driven by both structural consistency and physical rationality.

[0080] It should be understood that compared with the existing technology that only relies on surface point error regression or layer-by-layer error stacking, the present invention realizes continuous prediction of three-dimensional errors by introducing deep neural architecture and inter-layer fusion mechanism, and automatically obtains the error evolution law without manually setting rules, and significantly improves adaptability and prediction accuracy.

[0081] For example, in the error prediction task of a printed part with a highly complex topological structure, after using 800 sets of historical printing samples to train the model, the new two-dimensional error map (average layer stress 90MPa) and process data (laser power, scanning speed, material type) were input into the model as fusion vectors. The output three-dimensional error distribution had an average error of only 21μm compared with the actual measurement value on the test set. Compared with the traditional error stacking method (error of about 41μm), the prediction error was reduced by about 48.8%, the spatial coherence was stronger, and the accuracy distribution was closer to the actual thermal stress deformation trend.

[0082] Step S50: generating a print quality prediction report based on the spatial geometric accuracy level map and the target design dimensions of the target print part.

[0083] It should be noted that the generation process of the print quality prediction report includes: global accuracy level assessment, statistical voxel proportions of each level area in the spatial geometric accuracy level map; key structural area qualification judgment, extraction of structural key areas (such as hole positions, connection surfaces, precision surfaces, etc.) in the target print design drawing, and comparison of their corresponding error levels to determine whether they fall within the design tolerance range; report indicator generation: including maximum prediction error, root mean square error (RMSE), spatial error heat map, qualification rate index, prediction credibility score, etc.; the report supports export to structured files (such as PDF, JSON), which can be used as an important basis for printing task evaluation, process parameter review or quality traceability.

[0084] It should be understood that this print quality prediction report is not just a numerical statistical result; it is also a highly application-relevant result based on a structural combination of spatial grade information and design target dimensions. By matching it with the key dimensions and tolerances marked in the design model, it can provide an "upfront assessment" of print feasibility and expected achievability, helping users to predict potential failure areas before printing.

[0085] Embodiment 2: Furthermore, the present invention provides a metal 3D printing quality prediction system that employs a metal 3D printing quality prediction method described in the aforementioned embodiment, capable of resolving the technical problem of metal 3D printing quality prediction. Compared to the prior art, the metal 3D printing quality prediction system provided by the present invention has the same beneficial effects as the metal 3D printing quality prediction method described in the aforementioned embodiment. Other technical features of the metal 3D printing quality prediction system are the same as those disclosed in the aforementioned embodiment and are not further elaborated here.

[0086] Example 3: The present invention provides a metal 3D printing quality prediction device, please refer to Figure 2A metal 3D printing quality prediction device includes: at least one processor; and a memory in communication with the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute a metal 3D printing quality prediction method in the above-mentioned embodiment 1. A metal 3D printing quality prediction device in an embodiment of the present invention may include but is not limited to mobile terminals such as mobile phones, laptop computers, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Descriptions), PMPs (Portable Media Players), vehicle-mounted terminals (such as vehicle-mounted navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc. A metal 3D printing quality prediction device is merely an example and should not impose any limitations on the functions and scope of use of the embodiments of the present invention. A metal 3D printing quality prediction device may include a processing device 1001 (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes based on a program stored in a read-only memory (ROM) 1002 or a program loaded from a storage device 1003 into a random access memory (RAM) 1004. Various programs and data required for the operation of a metal 3D printing quality prediction device are also stored in RAM 1004. The processing device 1001, ROM 1002, and RAM 1004 are connected to each other via a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Typically, the following systems can be connected to the I / O interface 1006: input devices 1007 including, for example, a touch screen, touchpad, keyboard, mouse, image sensor, microphone, accelerometer, gyroscope, etc.; output devices 1008 including, for example, a liquid crystal display (LCD), speaker, vibrator, etc.; storage devices 1003 including, for example, a magnetic tape, hard disk, etc.; and communication devices 1009. The communication device 1009 can allow a metal 3D printing quality prediction device to communicate wirelessly or wired with other devices to exchange data. Although the figure shows a metal 3D printing quality prediction device with various systems, it should be understood that it is not required to implement or have all the systems shown. More or fewer systems can be implemented or have instead.

[0087] Example 4: The present invention also provides a computer program product, comprising a computer program. When executed by a processor, the computer program implements the steps of the aforementioned method for predicting metal 3D printing quality. The computer program product provided by the present invention can solve the technical problem of predicting metal 3D printing quality. Compared with the prior art, the beneficial effects of the computer program product provided by the present invention are the same as those of the method for predicting metal 3D printing quality provided by the aforementioned embodiment, and are not further elaborated here.

[0088] In particular, according to the embodiments disclosed in the present invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, the embodiments disclosed in the present invention include a computer program product comprising a computer program carried on a computer-readable medium, the computer program comprising program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network via a communication device, or installed from a storage device 1003, or installed from a ROM 1002. When the computer program is executed by the processing device 1001, the above-mentioned functions defined in the method of the embodiment disclosed in the present invention are performed.

[0089] It should be understood that the various parts disclosed in the present invention can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any appropriate manner in any one or more embodiments or examples.

[0090] Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if such changes and modifications fall within the scope of the claims and their equivalents, the present invention is intended to include such changes and modifications.

Claims

1. A metal 3D printing quality prediction method, characterized in that: Methods include: Step S10: collecting real-time status information of the 3D printer and current print status data, and calculating the current print thermal stress based on the real-time status information and the current print status data; Step S20: obtaining a target design size of a target printed part, and constructing a two-dimensional error map according to the target design size and the thermal stress of the current printed part; Step S30: The two-dimensional error map, real-time status information, and current print status data are normalized to obtain a fused input vector, a preset error distribution prediction model is pre-set, and the error distribution prediction model is fused and trained using a convolutional neural network and a time series network. The fused input vector is input into the trained error distribution prediction model to obtain a three-dimensional error distribution. Step S40: performing regional clustering and error level classification based on the three-dimensional error distribution to generate a spatial geometric accuracy level map of the target print; Step S50: generating a print quality prediction report based on the spatial geometric accuracy level map and the target design dimensions of the target print part.

2. A metal 3D printing quality prediction method according to claim 1, characterized in that: In step S10, the steps of collecting the real-time status information of the 3D printer and the current print status data, and calculating the thermal stress of the current print according to the real-time status information and the current print status data specifically include: Collect the real-time status information of the 3D printer and the current print status data, and obtain the melting zone temperature T at the time t position (x, y) from the current print status data melt (t, x, y), the thermal expansion coefficient α of the metal material and the elastic modulus E of the metal material, and obtain the printer cavity ambient temperature at the time t position (x, y) from the real-time status information; According to the melting zone temperature T melt (t, x, y) and the ambient temperature of the printer cavity are used to calculate the relative temperature difference; The global adjustment factor of the thermal stress of the current printed part is introduced, and the thermal stress of the current printed part is calculated based on the relative difference temperature, the global adjustment factor of the thermal stress of the current printed part, the thermal expansion coefficient α of the metal material, and the elastic modulus E of the metal material.

3. A metal 3D printing quality prediction method according to claim 2, characterized in that: In step S10, the real-time status information includes printer temperature, printer laser power and printer scanning speed; the current print status data includes metal material type, metal material thermal expansion coefficient, metal material elastic modulus, melting zone temperature and current print thermal stress.

4. A metal 3D printing quality prediction method according to claim 1, characterized in that: In step S30, the two-dimensional error map, real-time status information, and current print status data are normalized to obtain a fused input vector, a preset error distribution prediction model is preset, and the error distribution prediction model is fused and trained using a convolutional neural network and a time series network. The fused input vector is input into the trained error distribution prediction model to obtain a three-dimensional error distribution. The steps specifically include: Step S301: normalizing the two-dimensional error map, the real-time status information, and the current print status data to construct a fusion input vector in a unified numerical space; Step S302: using a convolutional neural network to extract local stress variation features and spatial correlation pattern features between geometric deformations from the fused input vector; Step S303: extracting temporal correlation pattern features during the printing process from the fused input vector using a time series network; Step S304: fusing the spatial correlation pattern features between the local stress variation features and the geometric deformation extracted by the convolutional neural network and the temporal correlation pattern features extracted by the time series network to obtain a feature fusion vector; Step S305: Preset the error distribution prediction model, obtain historical error distribution annotation data and historical feature fusion vector data, and use the historical error distribution annotation data and historical feature fusion vector data as input of the error distribution prediction model to pre-train the error distribution prediction model. Finally, pass the feature fusion vector through the inter-layer fusion layer and the fully connected layer of the pre-trained error distribution prediction model to output the three-dimensional error distribution.

5. A metal 3D printing quality prediction method according to claim 4, characterized in that: In step S30, the inter-layer fusion layer is used to achieve temporal fusion from the layered two-dimensional error map to the overall three-dimensional error distribution, wherein the formula of the inter-layer fusion layer is expressed as: in, is the predicted geometric error of the k-th layer position (x, y), where z k is the height coordinate value of the k-th layer position (x, y); ΔL (k) (x,y) is the two-dimensional error map of the k-th layer position (x,y), h k-1 is the hidden state passed through the k-1th layer of the error distribution prediction model, s k is the process parameter feature vector corresponding to the kth layer, f fusion (·) is the fusion function of the error distribution prediction model.

6. A metal 3D printing quality prediction method according to claim 1, characterized in that: In step S40, the steps of performing regional clustering and error level division according to the three-dimensional error distribution to generate a spatial geometric accuracy level map of the target print part specifically include: Step S401: performing voxel segmentation on the three-dimensional error distribution to obtain a plurality of spatial sub-regions, where the spatial sub-regions are used to represent the geometric deviation of each point of the target print in the three-dimensional space; Step S402: extracting local error statistical features of each spatial sub-region using a mathematical statistical method to form an error feature vector; Step S403: performing unsupervised cluster analysis on the multiple spatial sub-regions based on the error feature vector to obtain multiple cluster sub-regions; Step S404: obtaining a preset error level classification standard, performing level mapping on multiple cluster sub-regions according to the preset error level classification standard, and generating a spatial geometric accuracy level map of the target print.

7. A metal 3D printing quality prediction method according to claim 2, characterized in that: In step S20, the formula for the two-dimensional error map is expressed as: Where σ(t,x,y) is the thermal stress of the current printed part at position (x,y) at time t, ΔL(x,y) is a two-dimensional error map representing the linear geometric deformation at position (x,y), L0(x,y) is the target design dimension at position (x,y), E is the elastic modulus of the metal material, and δ is the global adjustment factor for linear geometric deformation.

8. A metal 3D printing quality prediction system, applied to a metal 3D printing quality prediction method according to any one of claims 1 to 7, characterized in that: The metal 3D printing quality prediction system includes: A state stress calculation module is used to collect real-time state information of the 3D printer and current print status data, and calculate the current print thermal stress based on the real-time state information and current print status data; An error mapping module is used to obtain the target design size of the target printed part and construct a two-dimensional error map based on the target design size and the thermal stress of the current printed part; The error prediction modeling module is used to obtain a fused input vector after normalizing the two-dimensional error map, real-time status information, and current print status data, preset an error distribution prediction model, and use a convolutional neural network and a time series network to train the error distribution prediction model. The fused input vector is input into the trained error distribution prediction model to obtain a three-dimensional error distribution; A grade determination module is used to perform regional clustering and error grade classification based on the three-dimensional error distribution, and generate a spatial geometric accuracy grade map of the target print; The quality prediction output module is used to generate a print quality prediction report based on the spatial geometric accuracy level map and the target design size of the target print part.

9. A metal 3D printing quality prediction device, characterized in that: The metal 3D printing quality prediction device includes: a memory, a processor, and a metal 3D printing quality prediction program stored in the memory and executable on the processor. When the metal 3D printing quality prediction program is executed by the processor, a metal 3D printing quality prediction method according to any one of claims 1 to 7 is implemented.

10. A computer program product, characterized in that The computer program product includes a metal 3D printing quality prediction program, which, when executed by a processor, implements a metal 3D printing quality prediction method according to any one of claims 1 to 7.

Images