Method for manufacturing hydraulic support column nest casting mold through 3D printing technology

Abstract

The invention discloses a method for manufacturing a hydraulic support column socket casting mold through the 3D printing technology, and relates to the technical field of mold manufacturing, the method comprises the following steps that a three-dimensional digital model of a column socket casting is obtained, and the model comprises a ball socket curved surface with the continuous change curvature; based on the geometrical characteristics of the ball socket curved surface, multiple sub-regions with different printing process requirements are identified, and differentiated printing process parameter sets are distributed to the sub-regions; printing a mold blank according to the printing process parameter set by adopting a fused deposition modeling process; and the mold blank is subjected to finish machining, and the casting mold is obtained. In the process of 3D printing of the column socket mold, the ball socket area is divided into a plurality of sub-areas of the high-risk area and the general area according to intelligent geometric analysis of the curvature and the normal direction, different printing strategies are adopted in the different sub-areas, the amplitude of the periodic wavy surface is weakened, and the size precision and the profile tolerance of the casting mold are remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of mold manufacturing technology, and in particular to a method for manufacturing hydraulic support column socket casting molds using 3D printing technology. Background Technology

[0002] Hydraulic supports are one of the core pieces of equipment in coal mining machinery. The critical load-bearing positions of each hydraulic support rely on castings, especially the hydraulic cylinders. The column socket is one of the important load-bearing components of the hydraulic support equipment. During the casting process of the column socket, because the casting itself has a ball socket that mates with the column head of the jack and the bottom of the cylinder, as well as a weight-reducing hole on its back, the molding process requires high-quality sand cores, strict requirements on worker skills, mold quality, and dimensional accuracy.

[0003] Existing molds generally use wooden molds to prepare sand cores. The most crucial part of wooden mold making is model assembly. Assembly requires operators to have strong blueprint reading skills and ingenious design and assembly abilities. Furthermore, staggered assembly requires cutting boards. Each board needs to be cut by workers using different woodworking machines. In order to ensure that there is not enough material after processing and to avoid repeated repairs, sufficient processing allowance must be ensured when staggering layers, that is, the smallest position can be processed. Therefore, the processing allowance of the model after cutting and assembling is very large, and the processing cycle is long.

[0004] Chinese patent application number CN201911248353.2 discloses a method for manufacturing a 3D printed sand mold. This invention eliminates the casting and processing steps of traditional metal molds by using 3D printing technology, and realizes the rapid manufacturing of molds for sand casting.

[0005] Although 3D printing technology can improve many of the limitations of traditional wooden molds, the fused deposition modeling process inevitably produces a physical phenomenon called the "surface step effect" when manufacturing objects with complex curved surfaces. The core functional surface of the column socket—the ball socket—is precisely a concave curved surface with a large curvature and extremely high requirements for continuity, making the step effect most pronounced here. Since the ball socket is the load-bearing mating surface of the hydraulic support column, its surface continuity, geometric accuracy, and roughness directly determine the stress distribution, sealing performance, and service life.

[0006] Furthermore, although the step effect can be mitigated by controlling the layer height of each layer during 3D printing, the large size of the mold and the diameter of the spherical cavity can reach hundreds of millimeters. In order to control the total printing time, it is impossible to use a layer height that is too small. If the layer height is too high, the step height difference and surface roughness accumulated on the curved surface of the large-sized spherical cavity will be magnified proportionally. Moreover, after the mold is printed, it needs to be finished by CNC equipment. CNC programming relies on an ideal and uniform blank allowance. Due to the existence of the step effect, it means that the actual blank surface is wavy. The effective allowance of the theoretical machining allowance is different at the peaks and troughs of the step effect. If this is not addressed, it is easy to have serious consequences of "overcutting" and "undercutting".

[0007] To address these issues, this invention proposes a method for fabricating hydraulic support column socket casting molds using 3D printing technology. Summary of the Invention

[0008] The purpose of this invention is to provide a method for manufacturing hydraulic support column socket casting molds using 3D printing technology, so as to solve the technical problems mentioned in the background art.

[0009] To achieve the above objectives, the present invention provides the following technical solution: a method for manufacturing a hydraulic support column socket casting mold using 3D printing technology, comprising the following steps:

[0010] A three-dimensional digital model of a column-and-socket casting is obtained, the model comprising a ball-and-socket surface with continuously varying curvature;

[0011] Based on the geometric features of the ball-and-socket surface and its corresponding casting process requirements, an analysis was conducted to identify multiple sub-regions with different printing process requirements, and a differentiated set of printing process parameters was assigned to each sub-region.

[0012] The mold blank is printed using a fused deposition modeling process based on the aforementioned printing process parameter set.

[0013] The mold blank is precision machined to obtain a casting mold.

[0014] Preferably, the sub-region identification includes curvature analysis and normal analysis of the ball-and-socket surface to identify high-risk areas that are prone to sand adhesion or dimensional deviations in castings during sand casting.

[0015] The allocation of differentiated printing process parameter sets to each sub-region includes:

[0016] Based on the results of the curvature analysis and normal analysis, a first set of printing process parameters is assigned to the high-risk area, and a second set of printing process parameters is assigned to other areas; wherein the layer height values ​​contained in the first set of printing process parameters are smaller than the layer height values ​​contained in the second set of printing process parameters.

[0017] Preferably, the high-risk area includes at least the area in the ball-and-socket surface where the curvature is greater than a first threshold and / or the angle between the normal and the mold-opening direction is greater than a second threshold.

[0018] Preferably, after assigning differentiated sets of printing process parameters to different regions, the method further includes the step of:

[0019] Identify the boundaries between adjacent sub-regions using different sets of printing process parameters;

[0020] A process buffer zone is planned at the intersection, and within the process buffer zone, the printing process parameters on both sides are smoothly and gradually transitioned to eliminate seams or textures on the mold surface that may be formed by abrupt parameter changes, which are not conducive to sand mold release.

[0021] Preferably, the width of the process buffer is dynamically planned based on the degree of curvature change on both sides of the boundary and / or the degree of influence of differences in printing process parameters on the surface morphology.

[0022] Preferably, within the process buffer, the gradual change of printing process parameters follows a preset non-linear transition curve.

[0023] Preferably, the identification of the sub-region and the generation of the printing process parameter set are performed based on a pre-built digital process knowledge base containing various column socket casting mold manufacturing experiences, specifically including:

[0024] Extract the feature parameters of the current 3D digital model;

[0025] Based on the aforementioned feature parameters, optimized process schemes that have been validated in production are matched and retrieved from the digital process knowledge base, or an initial process scheme is generated through a prediction model.

[0026] Preferably, the method further includes the step of:

[0027] Based on the aforementioned printing process parameter set, virtual printing simulation and mold-opening process simulation are performed to obtain virtual simulation prediction data;

[0028] During the printing of the mold blank, printing process data is collected in real time, and the actual morphological data of the mold blank is obtained after printing is completed.

[0029] The printing process data and actual morphology data are compared with the virtual simulation prediction data, and the simulation model used for the virtual printing simulation or the digital process knowledge base is calibrated based on the comparison results.

[0030] Preferably, the printing process data includes at least one of printhead temperature, extrusion pressure, and forming chamber temperature field distribution.

[0031] Preferably, the actual morphology data is obtained through three-dimensional scanning, and the comparison result includes a deviation distribution map between the actual morphology and the predicted morphology.

[0032] The beneficial effects of this invention are:

[0033] In the process of 3D printing a spherical socket mold, this invention divides the spherical socket area into multiple sub-regions, including high-risk and general regions, based on intelligent geometric analysis of curvature and normal. Different printing strategies are adopted in different sub-regions to reduce the amplitude of periodic wavy surfaces, resulting in stable load on the CNC tool during cutting. This avoids tool vibration, tool bounce, overcutting, or undercutting caused by drastic fluctuations in allowance, significantly improving the dimensional accuracy and contour of the casting mold. Furthermore, by identifying the boundaries between adjacent sub-regions and setting process buffers at these boundaries, this invention achieves a smooth, soft transition from one process state to another during printing. This minimizes problems such as material accumulation, wire drawing, or weakened bonding caused by sudden parameter changes, eliminating regular seams on the mold blank surface. Consequently, the surface continuity and consistency of the entire spherical socket surface are significantly improved, providing further assurance for obtaining defect-free sand mold cavities and high-quality castings. Attached Figure Description

[0034] Figure 1 This is a schematic diagram of the overall control process of a method for manufacturing hydraulic support column socket casting mold using 3D printing technology according to the present invention.

[0035] Figure 2 This is a schematic diagram of the timing control of the 3D printing process of the present invention. Detailed Implementation

[0036] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0037] Example 1

[0038] In existing technologies, 3D printing technology can improve many limitations of traditional wooden molds. However, when manufacturing objects with complex curved surfaces, the fused deposition modeling process inevitably produces a physical phenomenon called the "curved surface step effect". The core functional surface of the column socket - the ball socket - is a concave curved surface with a large curvature and extremely high continuity requirements, which makes the step effect most significant. The resulting periodic wavy surface will be directly replicated on the subsequent sand mold and casting, which is one of the main causes of sand adhesion defects on the surface of the casting.

[0039] Furthermore, although the step effect can be mitigated by controlling the layer height of each layer during 3D printing, the large size of the mold and the diameter of the ball socket can reach hundreds of millimeters. To control the total printing time, excessively small layer heights cannot be used. Excessively high layer heights will proportionally amplify the step height difference and surface roughness accumulated on the curved surface of the large-sized ball socket. After the mold is printed, it needs to be precision-machined by CNC equipment to obtain the casting mold. CNC programming relies on an ideal and uniform blank allowance. Due to the existence of the step effect, the actual blank surface is wavy, which causes the effective allowance of the theoretical machining allowance to be different at the peaks and troughs of the step effect. If it is not pre-processed, overcutting at the peaks of the step or undercutting at the troughs of the step are very likely to occur during CNC machining, which will seriously damage the dimensional accuracy and contour of the ball socket.

[0040] This embodiment was invented to solve the above problems.

[0041] Please see Figure 1 and Figure 2 As shown, an embodiment of the present invention provides a method for fabricating a hydraulic support column socket casting mold using 3D printing technology, comprising the following steps:

[0042] S100. Obtain a three-dimensional digital model of the column-and-socket casting, wherein the model includes a ball-and-socket surface with continuously varying curvature.

[0043] In this step, based on the two-dimensional engineering drawings of the target column socket casting, a three-dimensional model is created using three-dimensional software to generate a three-dimensional digital model containing complete geometric information. The three-dimensional digital model accurately expresses all the features of the column socket, wherein the core spherical socket forming surface in the column socket is a complex concave surface with continuously varying curvature.

[0044] It is important to note that during modeling, machining allowances should be added globally and evenly to all surfaces of the model that require machining, according to the casting process requirements. (For example This machining allowance is used for subsequent CNC finishing and grinding, and the machining allowance is... It must be greater than the maximum staircase effect peak height that may occur under unoptimized printing conditions, as estimated through theoretical calculations or preliminary experiments. .

[0045] S200. Based on the geometric features of the ball-and-socket surface and its corresponding casting process requirements, an analysis is performed to identify multiple sub-regions with different printing process requirements, and a differentiated set of printing process parameters is assigned to each sub-region.

[0046] Specifically, the aforementioned 3D digital model with machining allowance is imported into a software environment with curvature analysis capabilities. The software first discretizes the spherical-socket surface into a mesh, and then performs curvature analysis and normal analysis calculations for each mesh cell or path point. The specific steps are as follows:

[0047] Discretization of the spherical surface: Using the application programming interface of 3D modeling software or a general mesh processing library (such as Open3D, MeshLab), the spherical surface is divided into triangular meshes, and the mesh size is set according to the printing accuracy requirements (e.g., the side length is no more than 0.5mm).

[0048] Curvature analysis: For each mesh cell, calculate its Gaussian curvature or average curvature to quantify its local bending degree. The higher the curvature value, the more abrupt the surface turn is, and the more obvious the step effect is in the 3D printing process.

[0049] Normal analysis: Calculate the angle between the surface normal vector at this point and the printer's building direction (usually the Z-axis). This included angle directly reflects the degree of drooping in the area. The larger the included angle, the more difficult it is for the material to be deposited in mid-air during printing, and the more likely it is to droop or string.

[0050] Then, based on the calculation results of curvature analysis and normal analysis, logical criteria are set to automatically divide the entire spherical surface into regions with different printing process requirements. As a specific implementation method, the following threshold rules are set:

[0051] Curvature value greater than the set curvature threshold And / or the included angle of the normals is greater than the included angle threshold. The continuous curved surface area is identified as the first type of area. This type of area corresponds to the part of the ball socket where the curvature changes drastically (such as the side wall transition area) or the overhang angle is large (such as the edge). It is a "high-risk area" for step effect and printing defects and is most sensitive to printing quality. The area that does not meet the above conditions is identified as the second type of area. This type of area usually has a gentle curvature and a small overhang angle, and is relatively easy to print.

[0052] The above curvature threshold and included angle threshold The basis for this determination is: through preliminary printing tests, a correlation model was established between different curvature and included angle regions and the sand adhesion defect rate and dimensional deviation rate on the casting surface, and a critical value that can effectively distinguish high-risk defect regions was selected.

[0053] To facilitate understanding, the following are examples of its partitioning methods:

[0054] When the curvature value of a certain region is consistently greater than 0.05mm -1 and the included angle of the normal When the angle is greater than 45 degrees, the continuous area will be identified by the system as the first type of area. This type of area usually corresponds to the transition area of ​​the side wall of the ball socket and the edge of the opening. It is a "high-risk area" with a significant step effect and is prone to printing defects. Areas that do not meet the above threshold are identified as the second type of area, which usually corresponds to relatively flat parts such as the bottom of the ball socket.

[0055] After partitioning, different sets of printing process parameters are assigned to different types of areas. The core of this is the printing layer height: the first set of parameters is assigned to the first type of area, which uses a smaller layer height (e.g., 0.25mm-0.4mm), and is supplemented with a slower printing speed and optimized temperature settings to actively suppress the step height of the area and improve surface density.

[0056] A second set of parameters is assigned to the second type of region, which adopts a larger layer height (e.g., 0.5mm-0.8mm) and is matched with a higher printing speed, so as to improve the overall manufacturing efficiency while ensuring basic molding quality. This set of differentiated parameters is integrated into the final printing control command.

[0057] S300: Using fused deposition modeling (FDM) technology, the mold blank is printed according to the set of printing process parameters.

[0058] Specifically, based on the digital manufacturing instructions generated in step S200, which are bound to partitions and corresponding printing process parameter sets, a gantry-type 3D printing machine is controlled to perform printing operations. During the printing process, when printing to the first type of spherical cavity, it automatically switches to a low layer height, low speed and high precision mode; when printing to the second type of spherical cavity, it switches to a high layer height and high efficiency mode. In this way, a mold blank with an overall complete structure and pre-optimized surface quality of key curved areas (high-risk areas) is printed.

[0059] It is important to note that the printing material for the casting model must meet the following requirements:

[0060] 1. Meets the requirements of ordinary carbide cutting tools, does not stick to the tool during cutting, and has a minimum surface finish of Ra12.5 after machining;

[0061] 2. Non-hygroscopic, wear-resistant, Rockwell hardness above 100R;

[0062] 3. The overall cost is the same as that of wooden molds, and lower after mass production;

[0063] 4. It does not degrade, crack, or deform after processing;

[0064] 5. It can recycle waste materials and reuse them repeatedly.

[0065] Based on the above requirements, this embodiment selected PLA, PETC, nylon carbon fiber, nylon glass fiber, TPU and PP glass fiber for printing verification after comparing the performance of common 3D printing materials. Through actual printing and testing, it was found that PP glass fiber has the best processing performance, but the shrinkage rate is still too high. After setting a 5mm processing allowance in Cura programming, the processing allowance is insufficient due to material shrinkage, or some areas cannot be processed.

[0066] Therefore, in order to reduce the shrinkage rate, this embodiment improves the composition of the mold processing material. Specifically, based on the existing composition, the glass fiber content is adjusted, and in order to increase the surface smoothness, 15% ± 2% calcium carbonate is added, and the glass fiber addition amount is 20% ± 2%, to obtain the optimal shrinkage rate, i.e., 0.1% to 0.15%.

[0067] S400. The mold blank is precision machined to obtain a casting mold.

[0068] Specifically, the printed mold blank is clamped on a CNC machine tool. Since the ball-and-socket surface of the mold blank has been optimized, the step effect in the first type of region has been suppressed to the greatest extent. The wave undulation amplitude of the entire surface is significantly reduced and more uniform. Therefore, when the CNC tool cuts according to the preset ball-and-socket finishing toolpath, because the peak height of the actual surface of the blank (especially the critical first type of region) has been reduced and the valley depth has been increased, the actual effective value of the preset uniform machining allowance at any point on the surface becomes very close. This makes the load on the tool more stable when cutting, without the need for additional dynamic adjustments or vibrations due to the severe surface undulations.

[0069] More importantly, at the crests, the material will not be excessively removed due to excessively high actual protrusions, thus damaging the underlying structure; at the troughs, the material will not leave uncut "islands" due to excessively deep depressions; the machining process stably removes the preset allowance layer, and the final formed spherical cavity surface has precise dimensions (e.g., SR tolerance is stably controlled within 0 to +2 mm), high surface finish, and can be directly used for sand casting production, thereby improving the dimensional qualification rate of castings and reducing the risk of sand adhesion.

[0070] In summary, this invention, during the 3D printing of a socket mold, divides the socket area into multiple sub-regions, including high-risk and general regions, based on intelligent geometric analysis of curvature and normal. Different printing strategies are employed in different sub-regions to reduce the amplitude of the periodic wavy surface, resulting in a stable load on the CNC tool during cutting. This avoids tool vibration, tool bounce, overcutting, or undercutting caused by drastic fluctuations in allowance, significantly improving the dimensional accuracy and contour of the casting mold.

[0071] Furthermore, the optimized PP glass fiber composite material has an extremely low molding shrinkage rate, a characteristic that is crucial for the dimensional stability of large molds and effectively avoids insufficient processing allowance or deformation caused by shrinkage.

[0072] Example 2

[0073] In practical use, it was found that simply dividing the ball-and-socket surface for printing, although optimizing the quality and efficiency of different areas by allocating differentiated parameters, can also cause new defects at the junction of two adjacent areas during the printing process due to the sudden changes in extrusion flow, movement speed, and cooling conditions. This results in a visible seam, material accumulation band, or weak material bonding line on the surface of the mold blank. This regular texture introduced by the printing process itself will also be completely replicated on the surface of the sand mold cavity.

[0074] Therefore, this embodiment makes further improvements based on the above embodiments to address the above problems.

[0075] Please see Figure 1 As shown, after assigning differentiated sets of printing process parameters to different regions, the following steps are also included:

[0076] The system identifies the boundaries between adjacent planning areas that use different sets of printing process parameters. Specifically, after dividing the area according to the surface geometry and assigning differentiated parameter sets to each area, the system will automatically identify the theoretical boundary lines between all adjacent areas using different parameter sets.

[0077] A process buffer zone is planned at the intersection, and within the process buffer zone, the printing process parameters on both sides are smoothly and gradually transitioned to eliminate seams or textures on the mold surface that may be formed by abrupt parameter changes, which are not conducive to sand mold release.

[0078] The width of the process buffer is dynamically planned based on the degree of curvature change on both sides of the boundary and / or the degree of influence of differences in printing process parameters on the surface morphology.

[0079] It is important to note that the width W of the process buffer is not a fixed value, but is dynamically planned using a specific algorithm, as follows:

[0080] First, set a base buffer width that is related to the print nozzle diameter. ;

[0081]

[0082] in Where n is the nozzle diameter and n is an empirical coefficient that can be taken as 4 to 8.

[0083] Then, the difference in core printing parameters between the regions on both sides of the boundary line is calculated. In this embodiment, It's because of the poor printing speed. floor height difference Normalized weighted sum of key parameters.

[0084]

[0085] in, and These are the reference average values ​​for printing speed and floor height difference, respectively, and their weighting coefficients. , (If all are set to 0.5) This reflects the relative importance of each parameter to the interface quality, which can be preliminarily determined through sensitivity analysis.

[0086] Simultaneously, the degree of geometric abrupt change G near the boundary line is assessed; for example, the standard deviation of the average curvature of the surface within a certain range on both sides of the boundary line is calculated. Ultimately, the actual width of the process buffer zone is determined. Determined by the following formula:

[0087]

[0088] in, and These are calibration coefficients, and their calibration method is as follows: select several coefficients with different... and The typical boundary is determined by printing experiments, with the absence of visible seams at the boundary as the quality target, and the optimal α and β values ​​are derived (for example, by regression analysis to determine that they are in the range of 0.2-0.5 respectively).

[0089] The above calculation method ensures that a wider buffer is allocated at the boundary where there are large differences in parameters or drastic geometric changes, so as to provide sufficient transition space.

[0090] Within a defined process buffer, all printing parameters smoothly and continuously transition from values ​​in one region to values ​​in the other. This transition is not a simple linear interpolation, but rather controlled using a non-linear gradient curve that better matches the physical process. For example, the transition in printing speed is planned using an S-curve (Sigmoid function), the mathematical expression of which is approximately:

[0091]

[0092] In the formula:

[0093] This represents the target printing speed at normalized position t within the process buffer.

[0094] and The printing speed baseline values ​​set for the left and right adjacent areas respectively serve as the start and end points of the transition;

[0095] Represents the normalized position coordinates within the process buffer; it is a variable between 0 and 1. When, it represents the starting boundary of the buffer zone near the left side; when When, it represents the end boundary of the buffer zone near the right side; when This indicates that the location is exactly at the center line of the buffer zone (i.e., the original theoretical boundary line);

[0096] Represents the natural exponential function, that is, a mathematical constant Exponential operations with base 0, for example, That is This is the core mathematical operation that constitutes the S-curve;

[0097] The shape factor (k > 0) controls the smoothness of the transition curve and the speed at which the transition occurs; The larger the value, the closer the curve is to the center point. The steeper the change in the vicinity, the faster the transition occurs; The smaller the value, the smoother the curve is overall, and the transition occurs over a wider range; The value can be calibrated experimentally: choose a baseline boundary and try different... Print the value that will result in the most uniform fusion of the extruded filaments at the junction and the smoothest surface. The value is used as the calibration value for this material-equipment combination (e.g., ).

[0098] In summary, by identifying the boundaries between adjacent sub-regions and setting process buffer zones at these boundaries, this invention achieves a smooth, soft transition from one process state to another during the printing process. This minimizes problems such as material accumulation, wire drawing, or weakened bonding caused by sudden parameter changes, eliminates regular seams on the mold blank surface, and significantly improves the surface continuity and consistency of the entire spherical cavity surface. This provides further assurance for obtaining defect-free sand mold cavities and castings with high surface quality.

[0099] In summary, this invention solves the problems of material accumulation or weak bonding caused by abrupt changes in printing parameters during the printing process by setting process buffer zones between adjacent sub-regions and using a nonlinear parameter gradient algorithm based on S-shaped curves to convert hard switching of printing parameters into smoother soft switching. This improves or even avoids the formation of visible seams, material accumulation bands, or microscopic weak bonding surfaces at the boundaries of sub-regions, and enables the mold blank to achieve continuous and consistent surface texture across the entire ball-and-socket curved surface, thereby improving the stability of the mold release process during subsequent sand casting.

[0100] Furthermore, the width of the process buffer is not fixed, but dynamically planned through a specific algorithm that comprehensively considers the degree of geometric change and parameter differences. This ensures that a wider buffer is allocated to the areas that require the smoothest transition, and more transition resources are invested. In areas with gentle changes, a narrower buffer is used to reduce unnecessary printing paths. This adaptive mechanism reduces the additional impact on the overall printing efficiency caused by setting the process buffer while ensuring the quality of the interface, thereby maintaining the overall efficiency of the printing system.

[0101] Example 3

[0102] Furthermore, due to the wide variety of models and structures of column sockets, relying entirely on process engineers to manually perform surface analysis, zoning strategy design, and printing parameter debugging for each new model of column socket would result in huge time and trial-and-error costs. It would also lead to a long process preparation cycle, making it impossible to quickly respond to production needs. To a certain extent, this restricts the popularization and application efficiency of 3D printing technology in production scenarios with distinct characteristics of multi-variety and small-batch production.

[0103] Therefore, in the identification of the sub-region and the generation of the printing process parameter set in this embodiment, a digital process knowledge base based on data-driven and machine learning methods is introduced. This knowledge base serves as a core process decision engine, and its construction and application specifically include the following steps:

[0104] The identification of the sub-regions and the generation of the printing process parameter set are performed based on a pre-built digital process knowledge base containing various experiences in manufacturing column socket casting molds, specifically including:

[0105] Extract the feature parameters of the current 3D digital model;

[0106] Based on the aforementioned feature parameters, optimized process schemes that have been validated in production are matched and retrieved from the digital process knowledge base, or an initial process scheme is generated through a prediction model.

[0107] Specifically, a digital process knowledge base is constructed, which systematically stores a large number of successful printing cases that have been verified in production and correspond to different socket models. Each case is presented in the form of data pairs.

[0108] One is the feature vector, which is extracted from the original three-dimensional digital model of the corresponding model and standardized. The extracted features have clear physical and geometric meanings, including at least the global geometric features of the model (such as the maximum diameter, depth, volume, surface area of ​​the spherical socket, and the ratio of the overall volume of the casting to the volume of the spherical socket), statistical features describing the local geometric properties of the surface (such as the average curvature and the average, standard deviation, maximum value and distribution histogram of the average curvature and Gaussian curvature of all sampling points on the concave surface of the spherical socket), and features reflecting the topological morphology of the model (such as the shape description of the key cross-sections obtained based on the model skeleton analysis).

[0109] Secondly, there is a complete set of optimized process schemes that strictly correspond to the feature vector. This scheme not only includes the optimal partitioning rules verified for this model and the detailed printing parameters matched for each sub-region (including layer height, printing speed, nozzle temperature, etc.), but also includes the process buffer calculation coefficients (such as weighting coefficients α and β) determined to solve boundary problems and the shape factor k of the nonlinear transition curve.

[0110] Therefore, when it is necessary to print a new 3D digital model of a column socket When this happens, the system initiates the intelligent process decision-making process, which is as follows:

[0111] First, the system uses the above algorithm to extract... eigenvectors ,calculate Compared with all historical feature vectors in the knowledge base cosine similarity .

[0112] Then, the system uses a preset similarity threshold. Used to determine the confidence level of a match. The setting method is as follows: Analyze historical cases to determine the minimum similarity at which a process solution can be safely reused. Typically, this is taken as the lower quintile of the similarity distribution among all successful reuse cases (e.g., 0.85). If a highly similar case exists, its related process solution will be directly retrieved as... The solution.

[0113] like This indicates that the current column socket printing model is novel or has significant feature differences. The system will then use an embedded prediction model to generate an initial process plan. This prediction model is a gradient boosting decision tree regression model trained on all data in the knowledge base. The model takes feature vectors as input and various parameters in the process plan as output targets for multi-output regression. During training, mean squared error is used as the loss function, and cross-validation is used to prevent overfitting. The system will... After inputting the trained gradient boosting decision tree regression model, a complete set of initial process parameters can be predicted.

[0114] Furthermore, each time a new model is successfully validated in production, it is treated as a new case and incrementally updated to the training dataset of the digital process knowledge base and the gradient boosting decision tree regression model, enabling the system to learn continuously.

[0115] In summary, this invention constructs a digital process knowledge base containing numerous successful cases, enabling rapid feature matching and similarity retrieval of newly input column socket models. For models highly similar to historical models, mature process solutions that have been validated in production can be directly retrieved, achieving plug-and-play functionality. This shortens the process analysis and trial-and-error debugging process, which originally required several days or even weeks, to just a few minutes or even real-time, greatly reducing the process preparation time for new products and making flexible production of small batches and multiple varieties possible.

[0116] Furthermore, for highly innovative or significantly different new column socket models that are not present in the digital process knowledge base, the system can call a pre-trained machine learning model to generate a reasonable initial process plan based on the learned feature-parameter complex mapping relationship. This enables the printing system to have preliminary intelligence to cope with new challenges, no longer relying entirely on the precise matching of historical experience. Each newly completed and verified case will be fed back to the knowledge base, enabling the system to continuously learn and iterate its performance, thereby continuously enhancing its coverage and decision-making accuracy over time.

[0117] Example 4

[0118] In actual production, even if the sub-regional division scheme and digital process knowledge base are adopted in the above embodiments, there will still be unavoidable deviations between the printing results predicted by virtual simulation and physical reality due to fluctuations in material properties, equipment state drift, changes in environmental conditions, and inherent errors that may be caused by model simplification. If such deviations are not identified, quantified, and corrected, the optimization schemes in the process knowledge base will gradually deviate from the true optimal state, ultimately affecting the stability of mass production and the consistency of casting quality.

[0119] Therefore, based on the above embodiments, after printing the mold blank using the fused deposition modeling process, this embodiment also includes the following steps:

[0120] Based on the set of printing process parameters, virtual printing simulation and mold-opening process simulation are performed to obtain virtual simulation prediction data. The simulation is based on the finite element method. Furthermore, during the printing of the mold blank, printing process data is collected in real time, and the actual morphology data of the mold blank is obtained after printing is completed.

[0121] The above printing process data and actual morphology data are compared with the virtual simulation prediction data, and the simulation model or the digital process knowledge base used for the virtual printing simulation is calibrated based on the comparison results.

[0122] The aforementioned printing process data includes at least one of printhead temperature, extrusion pressure, and forming chamber temperature field distribution; the aforementioned actual morphology data is obtained through three-dimensional scanning; and the aforementioned comparison results include a deviation distribution map between the actual morphology and the predicted morphology.

[0123] Specifically, after determining the complete printing strategy (including partitioning scheme, printing process parameter set, and boundary transition rules) for the current model of column socket based on the digital process knowledge base, the system will initiate a high-fidelity virtual printing simulation. Based on the finite element simulation method, it constructs a simplified thermo-mechanical sequential coupling model, which includes:

[0124] Thermal process simulation: Using the printing path and time as the heat source movement trajectory, the three-dimensional transient heat conduction equation is solved to predict the historical distribution of the temperature field during and after printing. The thermal properties of the material are obtained from the measured database.

[0125] Deformation and stress simulation: Based on the above temperature field results, the residual stress and deformation caused by non-uniform cooling are calculated by using a thermo-elastic-plastic model to predict the theoretical three-dimensional morphology of the mold blank.

[0126] Simulation of the mold release process: Based on the theoretical morphology, the sand mold is applied to the mold with a clamping force. Through static finite element analysis, high-risk areas of stress concentration or potential interference during mold release are identified.

[0127] The simulation results above are integrated into virtual simulation prediction data, which includes temperature-time curves, three-dimensional topographic meshes, and high-risk area markers.

[0128] Subsequently, the actual printing stage begins. During this process, the system synchronously collects printing process data in real time through a multi-source sensor network integrated into the gantry 3D printer. The printing process data includes at least: the actual temperature curve of the print head monitored in real time by thermocouples embedded in the print head heating block; dynamic data of the extrusion pressure inside the nozzle monitored by melt pressure sensors; and time-series data of the three-dimensional temperature field distribution in the forming chamber monitored by infrared temperature sensors or thermocouple arrays placed at multiple key locations in the forming chamber (such as near the center, four corners, and top of the printing platform). All process data are synchronously collected, timestamped, and stored at a frequency of no less than 1Hz, forming a complete digital twin record of the print.

[0129] After printing, the mold blank needs to be moved to a testing station equipped with a high-precision blue light or laser 3D scanner. The scanner performs a full-range scan of the mold blank, especially the ball socket area, to obtain point cloud data of its actual outer surface. The point cloud data is then processed for noise reduction, alignment, and encapsulation to generate a high-resolution actual 3D shape model. Quantitative actual shape data such as actual contour dimensions and surface roughness distribution map are extracted from the model.

[0130] Secondly, the system compares and analyzes the deviations between the virtual simulation prediction data and the actual topographic data obtained above in the same spatiotemporal coordinate system:

[0131] Process data comparison: Align the measured temperature and pressure curves with the simulation prediction curves, and calculate the absolute deviation and root mean square error (RMSE) at each time point.

[0132] Topographic data comparison: The measured 3D model and the simulated predicted 3D model are compared in 3D. The normal distance of corresponding point pairs within the entire model and the partition range is calculated, and a color deviation cloud map is generated. In this map, different colors are used to clearly mark the areas where the actual size is greater than the predicted value (positive deviation is displayed in warm colors), the areas where the actual size is less than the predicted value (negative deviation is displayed in cool colors), and the areas that meet the prediction (near zero deviation). The system will automatically calculate quantitative indicators such as the maximum value, average value, standard deviation, and root mean square error of the global and key sub-regions.

[0133] To determine whether calibration is required, the system calculates the following deviation thresholds either by pre-setting or dynamically:

[0134] Process data deviation threshold: Set the upper limit of allowable process deviation for the root mean square error of temperature and pressure. For example, the root mean square error of temperature is ≤5℃ and the root mean square error of pressure is ≤0.5MPa.

[0135] Topographic data deviation threshold: Set the upper limit of allowable topographic deviation for the overall and key area size deviations. For example, the average absolute deviation of the key area is ≤0.1mm and the maximum absolute deviation is ≤0.3mm.

[0136] Finally, based on the comparison results of the above process data domain morphology data, the system determines whether the simulation model or digital process knowledge base needs to be calibrated. The specific process is as follows:

[0137] The system will compare the various deviation indicators obtained from the comparison calculation with the preset deviation thresholds mentioned above.

[0138] If all deviation indicators are within the threshold range, the printing result is considered to meet expectations, the current process is effective, no calibration is required, and the relevant data can be stored in the knowledge base as a case study to enrich the sample.

[0139] If one or more deviation indicators exceed their corresponding deviation thresholds, it is determined that the current process plan deviates significantly from the actual production status, triggering the calibration process. The specific calibration method is as follows:

[0140] Simulation model parameter calibration: If the deviation analysis shows that the RMSE of the temperature or pressure curve in the actual printing process exceeds the upper limit of the process deviation and exhibits systematic deviation characteristics, the system will start the reverse optimization algorithm. For example, if the actual printhead temperature is generally lower than the simulation set value, the algorithm will automatically adjust the thermal conductivity coefficient of the material in the simulation model or the heat exchange coefficient with the environment, so that the temperature prediction of the next simulation is closer to reality.

[0141] Digital process knowledge base calibration: If the deviation of the actual morphology data (especially the deviation of key areas) exceeds the upper limit of morphology deviation, or if the current practice does not exceed the tolerance but obtains a better result than the original scheme through fine-tuning, the system will start knowledge base calibration; that is, the system will include the original 3D model features, the applied process strategy, the collected actual process data, the final actual morphology deviation data, and the complete data of casting quality feedback as a complete parameter pair, and perform correlation analysis with the original case of the process scheme retrieved in the knowledge base.

[0142] If this practice yields better results than the original solution, or if shortcomings of the original solution are discovered, the system will, after confirmation by the engineer, create a new version of the process solution for that model, or generate a process correction rule for specific characteristics, and update it to the digital process knowledge base.

[0143] Meanwhile, the model feature weights used for similarity matching in the digital process knowledge base can also be adaptively adjusted based on the deviation feedback from a large number of cases, making them pay more attention to those geometric features that have a greater impact on the final print quality.

[0144] In summary, this invention quantitatively compares the real-time collected actual mold morphology data with the virtual simulation prediction results, and automatically optimizes the simulation parameters based on the deviation, thus constructing a complete closed-loop feedback system. This system can continuously monitor and compensate for systematic deviations caused by material fluctuations, equipment state drift, etc. For example, when a trend change in extrusion pressure is detected, the system can issue an early warning or automatically calibrate relevant parameters, thereby preventing quality problems from occurring.

[0145] Furthermore, if the optimized solutions in the digital process knowledge base are not verified against real production data over a long period of time, they may gradually become ineffective due to equipment wear and tear and differences in material batches. This invention feeds back the complete data chain of each production to the knowledge base and analyzes it in relation to the original solution. It can automatically discover and correct the shortcomings of the original solution or create a better new version. This makes the knowledge base no longer a static archive, but a knowledge system with self-correction and growth capabilities, ensuring the long-term effectiveness of its decisions.

[0146] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for fabricating a hydraulic support column socket casting mold using 3D printing technology, characterized in that, Includes the following steps: A three-dimensional digital model of a column-and-socket casting is obtained, the model comprising a ball-and-socket surface with continuously varying curvature; Based on the geometric features of the ball-and-socket surface and its corresponding casting process requirements, an analysis was conducted to identify multiple sub-regions with different printing process requirements, and a differentiated set of printing process parameters was assigned to each sub-region. The mold blank is printed using a fused deposition modeling process based on the aforementioned printing process parameter set. The mold blank is precision machined to obtain a casting mold.

2. The method for fabricating a hydraulic support column socket casting mold using 3D printing technology according to claim 1, characterized in that, The sub-region identification includes curvature analysis and normal analysis of the ball-and-socket surface to identify high-risk areas that are prone to sand adhesion or dimensional deviations in sand casting. The allocation of differentiated printing process parameter sets to each sub-region includes: Based on the results of the curvature analysis and normal analysis, a first set of printing process parameters is assigned to the high-risk area, and a second set of printing process parameters is assigned to other areas; wherein the layer height values ​​contained in the first set of printing process parameters are smaller than the layer height values ​​contained in the second set of printing process parameters.

3. A method for fabricating a hydraulic support column socket casting mold using 3D printing technology according to claim 2, characterized in that, The high-risk area includes at least the region in the ball-and-socket surface where the curvature is greater than a first threshold and / or the angle between the normal and the mold-opening direction is greater than a second threshold.

4. A method for fabricating a hydraulic support column socket casting mold using 3D printing technology according to claim 1, characterized in that, After assigning differentiated sets of printing process parameters to different regions, the following steps are also included: Identify the boundaries between adjacent sub-regions using different sets of printing process parameters; A process buffer zone is planned at the intersection, and within the process buffer zone, the printing process parameters on both sides are smoothly and gradually transitioned to eliminate seams or textures on the mold surface that may be formed by abrupt parameter changes, which are detrimental to sand mold release.

5. A method for fabricating a hydraulic support column socket casting mold using 3D printing technology according to claim 4, characterized in that, The width of the process buffer is dynamically planned based on the degree of curvature change on both sides of the boundary and / or the degree of influence of differences in printing process parameters on the surface morphology.

6. A method for fabricating a hydraulic support column socket casting mold using 3D printing technology according to claim 4, characterized in that, Within the process buffer, the gradual change of printing process parameters follows a preset non-linear transition curve.

7. A method for fabricating a hydraulic support column socket casting mold using 3D printing technology according to claim 1, characterized in that, The identification of the sub-regions and the generation of the printing process parameter set are performed based on a pre-built digital process knowledge base containing various experiences in manufacturing column socket casting molds, specifically including: Extract the feature parameters of the current 3D digital model; Based on the aforementioned feature parameters, optimized process schemes that have been validated in production are matched and retrieved from the digital process knowledge base, or an initial process scheme is generated through a prediction model.

8. A method for fabricating a hydraulic support column socket casting mold using 3D printing technology according to claim 1 or 7, characterized in that, It also includes the following steps: Based on the aforementioned printing process parameter set, virtual printing simulation and mold-opening process simulation are performed to obtain virtual simulation prediction data; During the printing of the mold blank, printing process data is collected in real time, and the actual morphological data of the mold blank is obtained after printing is completed. The printing process data and actual morphology data are compared with the virtual simulation prediction data, and the simulation model used for the virtual printing simulation or the digital process knowledge base is calibrated based on the comparison results.

9. A method for fabricating a hydraulic support column socket casting mold using 3D printing technology according to claim 8, characterized in that, The printing process data includes at least one of the following: printhead temperature, extrusion pressure, and temperature field distribution in the forming chamber.

10. A method for fabricating a hydraulic support column socket casting mold using 3D printing technology according to claim 8, characterized in that, The actual morphological data is obtained through three-dimensional scanning, and the comparison results include a deviation distribution map between the actual morphology and the predicted morphology.

Images