AI-assisted laser tilt engraving method for preparing complex three-dimensional microstructure and application thereof
By employing an AI-powered laser tilting engraving method and utilizing low-power multiple scanning and iterative optimization techniques, the problem of manufacturing sharp ends and smooth tilted sidewalls, which are difficult to achieve in traditional laser processing, has been solved, enabling the fabrication of complex three-dimensional microstructures with high precision.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-03-10
- Publication Date
- 2026-06-16
Smart Images
Figure CN122210237A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of micro-nano structure fabrication technology, and in particular to an AI-enabled laser tilting engraving method for preparing complex three-dimensional microstructures and its applications. Background Technology
[0002] Laser processing technology, especially laser marking / engraving technology based on galvanometer scanning, is widely used in the micro-processing of materials such as plastics, metals, and ceramics due to its advantages such as flexible processing, high efficiency, and non-contact operation. Its core principle is to guide the laser beam onto a focusing field lens (F-Theta Lens) by computer-controlled deflection of two galvanometers, forming a uniform, high-speed movable focused spot on its focal plane, thereby achieving selective removal (ablation) or modification of the material.
[0003] In fields such as biomedical engineering, microfluidic chips, micro-optics, and advanced sensors, the demand for micro / nanostructures with specific three-dimensional morphologies (such as continuous smooth inclined sidewalls and sharp ends) is increasingly urgent. For example, microneedle arrays are used for painless transdermal drug delivery, and their drug delivery efficiency is closely related to the sharpness of the needle tip and the taper of the needle body. The efficiency of mixers in microfluidic chips depends on the cross-sectional shape of the microchannels; V-grooves can significantly improve the mixing effect by inducing eddies. The performance of micro-optical elements (such as microlens arrays) and biomimetic functional surfaces (such as superhydrophobic surfaces) is directly determined by the accuracy and smoothness of their three-dimensional contours.
[0004] However, the fundamental design goal of traditional laser processing technology is to obtain processing conditions with a consistent focal spot size, uniform energy density, and perpendicular beam incidence throughout the entire processing plane, as well as to achieve two-dimensional or quasi-three-dimensional structures with vertical sidewalls and clear contours throughout the processing area. This paradigm severely limits its application in the fabrication of the aforementioned complex three-dimensional micro / nano structures.
[0005] To overcome the limitations of traditional two-dimensional laser processing and achieve complex three-dimensional structures, the commonly used method is layer-by-layer scanning engraving (also known as "3D laser engraving"). This method first "slices" the target 3D model along the Z-axis into a series of discrete two-dimensional contour layers. Then, by changing the laser processing parameters (such as focal height and laser power) layer by layer and performing two-dimensional scanning, the layers are stacked to eventually approximate a three-dimensional structure. However, for structures requiring very sharp ends (such as microneedle tips), the interlayer residue and thermal effects of the layer-by-layer method can cause the tips to become blunt, limiting the accuracy. Summary of the Invention
[0006] The purpose of this invention is to provide an AI-enabled laser tilting engraving method for preparing complex three-dimensional microstructures and its application, which utilizes the beam tilting and focusing effect of laser to control the morphology of the three-dimensional structure.
[0007] To achieve the above objectives, this invention provides a method for fabricating complex three-dimensional microstructures using AI-enabled laser tilting engraving, comprising the following steps: S1. Establish a process database; S2. Input the data from the process database into the forward prediction model for training to improve the accuracy and efficiency of the forward prediction model; S3. Based on the requirements of the three-dimensional structure, use the forward prediction model to reverse-engineer the marking parameters of the two-dimensional static pattern. S4. Position the workpiece to be processed and set the two-dimensional static pattern at the target position; S5. Perform low-power multiple scanning processing on the two-dimensional static pattern to obtain the initial three-dimensional structure; S6. Iteratively optimize the three-dimensional structure initially formed in step S5.
[0008] Preferably, step S1 specifically includes: S11. In the laser processing software, create a series of simple static test patterns and place the static test patterns at different coordinate positions within the scanning field; the different coordinate positions cover various areas from the center of the scanning field to the edge of the scanning field. S12. Set a set of fixed laser parameters and scan the static test pattern at each coordinate position in step S11 N times. S13. Using a surface profilometer or confocal microscope, accurately measure the three-dimensional morphological features of the grooves processed at each location; the three-dimensional morphological features are used to describe the complete shape information of the micro / nano structure in three-dimensional space, including the three-dimensional contour, shape changes and spatial positional relationships of the entire structure; S14. Associate the three-dimensional morphological features with the field coordinates of each static test pattern to form a mapping table or relationship curve, and construct the corresponding process database.
[0009] Preferably, step S2 specifically includes: S21. Perform data cleaning and feature engineering on the data in the process database; S22. Construct and train the positive prediction model using a deep neural network, and perform multiple verifications; S23. The positive prediction model is frozen using a genetic algorithm, and the input training data is continuously adjusted through an optimization algorithm until the three-dimensional structure data output by the positive prediction model is close to the actual three-dimensional structure data.
[0010] Preferably, step S3 specifically includes: S31. The user determines the three-dimensional structural data of the target three-dimensional structure according to the needs of the final product. S32. Based on the target three-dimensional structure, design a corresponding two-dimensional static pattern; S33. Based on the trained positive prediction model, obtain the optimal marking parameters corresponding to the target three-dimensional structure.
[0011] Preferably, step S33 includes: randomly generating a set of marking parameters; inputting the marking parameters into a forward prediction model to predict the three-dimensional structure data; when the predicted three-dimensional structure data does not match the three-dimensional structure data of the target three-dimensional structure, adjusting the marking parameter values using an optimization algorithm; repeating the iteration until the optimal marking parameters are found; and inputting the obtained optimal marking parameters into the forward prediction model for verification.
[0012] Preferably, step S4 specifically includes: S41. Fix the workpiece to be processed flat on the laser processing platform; S42. In the laser control software, drag and drop or position the two-dimensional static pattern to the target position at the edge of the scanning field obtained in step S33.
[0013] Preferably, step S5 specifically includes: S51. Set the laser power to a relatively low level and set the number of scans. The relatively low level is such that excessive energy in a single scan can prevent the material from undergoing severe vaporization and thermal damage. S52. Control the laser beam to repeatedly scan the two-dimensional static pattern that has been positioned at the target location; S53. Under the low-power multi-scan process, excess material is removed in a small amount and gradually to obtain the formed three-dimensional structure.
[0014] Preferably, in step S53, the two-dimensional static pattern located at the edge of the scanning field, since each point of the two-dimensional static pattern is irradiated by the tilted beam, the common area of the tilted beam inside the material naturally forms a tilted, converging three-dimensional structure.
[0015] Preferably, step S6 specifically includes: S61. After completing the scanning process, replace the scanning pattern with a smaller and finer outline in the laser control software, and align it precisely with the previously processed target position. S62. The scanning pattern is scanned a small number of times with the same or lower power to obtain the final shaped three-dimensional structure.
[0016] This invention also provides an application of the laser-engraved micro / nano structures obtained by the AI-enabled laser tilting engraving method described above in the fields of biomedical devices, microfluidic chips, micro-optical elements, and functional surface technologies.
[0017] Therefore, the present invention adopts the above-mentioned AI-enabled laser tilting engraving method for preparing complex three-dimensional microstructures and its application, which has the following beneficial effects: by adopting a strategy of multiple scans with low laser power, while avoiding excessive ablation and thermal damage to the material, it accurately and progressively shapes sharp ends and smooth tilted sidewalls, thereby overcoming the problem of insufficient precision.
[0018] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description
[0019] Figure 1 A flowchart illustrating an AI-enabled laser tilting engraving method for fabricating complex three-dimensional microstructures, according to an embodiment of the present invention. Figure 2 This is a schematic diagram of the core structure of a conventional laser scanning processing system according to an embodiment of the present invention; Figure 3 These are schematic diagrams illustrating the forming of various shapes of laser-engraved micro / nano structures according to embodiments of the present invention; Figure 4 This is a schematic diagram of superimposing a two-dimensional static pattern on a shaped three-dimensional structure according to Embodiment 2 of the present invention; Figure 5 This is a schematic diagram of the molding process of a three-dimensional structure after superimposing a two-dimensional static pattern on it, according to Embodiment 2 of the present invention. Detailed Implementation
[0020] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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, not all, of the embodiments of the present invention. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.
[0021] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.
[0022] Example 1 like Figure 1 As shown, this invention provides a method for fabricating complex three-dimensional microstructures using AI-enabled laser tilting engraving, comprising the following steps: S1. Establish a process database, which is the foundation for achieving precise control and repeatability.
[0023] In this embodiment, step S1 specifically includes: S11. In the laser processing software, create a series of simple static test patterns (such as circular dots with a diameter of 100 micrometers) and place them at different coordinate positions within the scanning field, especially covering the various areas from the center to the edge.
[0024] S12. Set a set of fixed laser parameters (e.g., a low power P0, scanning speed V0), and perform N (e.g., 5) scans on the static test pattern at each of the above coordinate positions.
[0025] S13. Using a surface profilometer or confocal microscope, accurately measure the three-dimensional morphological features of the grooves processed at each location. The three-dimensional morphological features are used to describe the complete shape information of the micro / nano structure in three-dimensional space, not limited to local angles or depths, but including the three-dimensional contour, shape changes, and spatial relationships of the entire structure.
[0026] S14. Associate the three-dimensional morphological features with the field coordinates (X, Y) of each static test pattern to form a mapping table or relationship curve, and construct the corresponding process database. This process database will clearly reveal that the closer to the edge of the scanning field, the smaller (sharper) the bottom tilt angle α of the processed structure.
[0027] S2. Input the data from the process database (equivalent to marking parameters) into the forward prediction model for training to improve the accuracy and efficiency of the forward prediction model.
[0028] In this embodiment, step S2 includes: S21. Perform data cleaning and feature engineering on the data in the process database (i.e., control the numerical range of the data between [0, 1]).
[0029] S22. Use a deep neural network (DNN) to build and train a forward prediction model, and perform multiple validations to ensure that the model's prediction accuracy is very high, so as to guarantee the accuracy of subsequent reverse predictions.
[0030] Training data (i.e., marking parameters) includes board thickness, number of markings, power, two-dimensional static pattern coordinates (x, y), marking speed, air jump speed, and Q frequency. Three-dimensional structural data (data that can clearly describe the three-dimensional structure) includes, but is not limited to, curvature, indentation height, tilt angle, indentation angle, accuracy, and shape.
[0031] S23. Use genetic algorithms to freeze the positive prediction model, and continuously adjust the input training data through the algorithm (Bayesian optimization) until the three-dimensional structure data output by the positive prediction model is close to the actual three-dimensional structure data.
[0032] The forward prediction model is a DNN model, consisting of an input layer, hidden layers, and an output layer. The input layer receives the raw data and passes data between the input layer and the hidden layers via a fully connected loop. The hidden layers abstract the input data (e.g., standardize numerical features) and perform nonlinear transformations. The hidden layers also pass data between themselves via a fully connected loop, use ReLU activation (to alleviate gradient vanishing and improve computational efficiency), and use Dropout for random deactivation (to prevent overfitting). The output layer generates the prediction results, and the hidden layers pass data between themselves via a fully connected loop.
[0033] Training iterations for the positive prediction model: With 1000 samples and 5000 features, when the batch size is 20, one epoch requires 250 iterations, meaning one iteration is completed after processing 20 samples.
[0034] The convergence condition for a positive prediction model is as follows: training terminates when the training loss decreases by less than a threshold or fluctuates within a very small range for multiple consecutive rounds, or when the loss / accuracy on the validation set no longer improves. The positive prediction model's performance has reached its optimal level when the loss function stabilizes.
[0035] S3. Based on the requirements of the three-dimensional structure, use the forward prediction model to calculate the marking parameters of its two-dimensional static pattern.
[0036] In this embodiment, step S3 specifically includes: S31. The user determines the three-dimensional structural data of the target three-dimensional structure according to the needs of the final product. For example, it is necessary to process a triangular pyramidal groove for a microneedle mold.
[0037] S32. Based on the target's three-dimensional structure, design the corresponding two-dimensional static pattern. For example, to obtain a triangular pyramid, design a triangular pattern; to obtain a square pyramid, design a square pattern.
[0038] S33. Based on the trained forward prediction model, obtain the optimal labeling parameters corresponding to the target 3D structure. Specifically, generate a set of labeling parameters randomly and input the labeling parameters into the forward prediction model to predict the 3D structure data. When the predicted 3D structure data does not match the 3D structure data of the target 3D structure, adjust the labeling parameter values using an optimization algorithm (such as Bayesian optimization). Since the labeling parameters contain multiple variables, one variable is used as a fixed value during adjustment. Repeat the iteration until the optimal labeling parameters are found. Input the obtained optimal labeling parameters into the forward prediction model for verification.
[0039] S4. Position the workpiece to be processed and set the two-dimensional static pattern at the target position.
[0040] In this embodiment, step S4 specifically includes: S41. Fix the workpiece to be processed (such as a PMMA board) flat on the laser processing platform (such as a Han's Laser CO2 laser marking machine).
[0041] S42. In the laser control software, drag and drop or position the two-dimensional static pattern (such as a triangle) designed in step S32 to the target position at the edge of the scanning field calculated in step S33. This step only involves setting the pattern position and does not require adjusting the physical Z-axis height of the device.
[0042] S5. Perform low-power multiple scanning processing on the two-dimensional static pattern to obtain the initial three-dimensional structure. This is a key process step to achieve high-quality molding.
[0043] In this embodiment, step S5 specifically includes: S51. Set the laser power to a relatively low level (to avoid excessive energy in a single scan leading to severe material vaporization and thermal damage), and set the number of scans (N times, N≥2). For example, when using a Han's Laser CO2 laser marking machine, set the laser power to 5%.
[0044] S52. Control the laser beam to repeatedly scan the two-dimensional static pattern that has been positioned at the target location.
[0045] S53. Under a low-power, multi-scan process, excess material is removed gradually in small amounts to obtain a shaped three-dimensional structure. The two-dimensional static pattern located at the edge of the scanning field, because each point is illuminated by an inclined beam of light, naturally forms an inclined, converging three-dimensional structure within the material due to the combined effect of these inclined beams. The generated three-dimensional structures are not limited to simple cones. For example, a triangular pattern will "grow" into a triangular pyramidal groove, a star shape will produce a star cone, a cross shape will produce a cross cone, and any polygon will produce a corresponding polygonal pyramid. This process does not require complex dynamic paths; the three-dimensional structure can be formed simply through repeated scanning of the two-dimensional static pattern.
[0046] S6. Iterative optimization of the three-dimensional structure initially formed in step S5 can locally refine the structure, further sharpen the tip, and optimize the smoothness of the sidewalls to achieve higher precision processing.
[0047] In this embodiment, step S6 specifically includes: S61. After completing the scanning process, replace the scanning pattern with a smaller and finer outline in the laser control software (for example, the original pattern is a right triangle with two right angles of 1mm each, and the reduced pattern is a right triangle with two right angles of 0.7mm each). (For example, a triangle with an area reduced by 50%) and align it precisely with the previously processed target position.
[0048] S62. Using the same or lower power (for example, setting the power of the Han's Laser CO2 laser marking machine to 5% or 4%), scan the pattern a small number of times (e.g., 2-3 times) to obtain the final formed three-dimensional structure, refer to... Figure 3 This step essentially involves controlling the energy input in a single pulse to achieve a gradual shaping strategy. This step can also be achieved through various parameter combinations, such as using a higher scanning speed to shorten the laser interaction time, or employing a high-frequency, low-energy pulsed laser. The core principle is to avoid excessively high energy in a single pulse, which could lead to uncontrollable and violent vaporization.
[0049] The processing equipment used in this invention includes laser equipment that utilizes focusing lenses for processing, such as CO2 laser marking machines and fiber laser marking machines. The materials to be processed include PMMA (acrylic), other plastics, polymer films, and even metals and ceramics (under suitable laser parameters). This method is also applicable to other polymers that absorb specific wavelengths of laser light well (such as polycarbonate PC and polyimide PI), and can even be used to create functional microstructures on metal surface coatings, wood, leather, and other materials.
[0050] This invention controls the morphology of a three-dimensional structure (especially the sidewall tilt angle and end sharpness) by controlling the position of the scanning pattern in the scanning field and actively utilizing the inherent beam tilting and converging effects. It provides an iteratively optimized process that allows for localized refinement and surface smoothing of the structural contour by replacing or superimposing finer scanning patterns after scanning based on the same principle, thereby further improving the shape accuracy and quality of the final structure.
[0051] This invention eliminates the need for complex 3D model slicing and repetitive Z-axis movements and scans, significantly reducing process complexity. Machining a triangular pyramid tip requires only repeatedly scanning a simple static triangular pattern at the edge of the scanning field. This new paradigm, prioritizing "Where" over "How," makes operation extremely simple and intuitive, greatly reducing the difficulty and time required for process development.
[0052] This invention can directly process well-defined, regular three-dimensional geometric shapes (such as obtaining a regular triangular pyramid by scanning a triangle at its edge). This is an inevitable result of utilizing the natural physical properties of light beams, and therefore has extremely high repeatability and stability.
[0053] This invention also provides an application of the laser-etched micro / nanostructures obtained by the AI-enabled laser tilting engraving method for fabricating complex three-dimensional microstructures, as described above, in the fields of biomedical devices, microfluidic chips, micro-optical elements, and functional surface technologies. The core of this application is the use of a conventional laser scanning system (see [reference]). Figure 2 By placing specific static patterns at the edge of the scanning field and employing a low-power, multiple-scan process, micro- and nanostructures with sharp ends, smooth, sloping sidewalls, and specific three-dimensional morphologies can be fabricated at low cost and high efficiency.
[0054] In the field of biomedical device technology, specific applications include, but are not limited to, the following products: Microneedle array mold: A mold used to fabricate polymer microneedles for applications such as painless transdermal drug delivery, vaccine injection, and tissue fluid extraction. Static patterns such as triangles, squares, or circles are placed at the edge of the scanning field. Through multiple low-power scans, microneedle pit molds with sharp tips and smooth sidewalls are directly fabricated on materials such as PMMA. By merging patterns and optimizing cross-scanning, the sharpness and shape of the needles can be precisely controlled.
[0055] Tissue engineering scaffolds: These are cell growth scaffolds with guided microstructures used in tissue engineering fields such as nerve regeneration and muscle repair. By fabricating V-shaped or U-shaped microgroove arrays on the surface of the scaffold material, the "contact guidance" effect of cells is utilized to guide cell growth in a specific direction. Complex guiding networks can be constructed through combinations of different patterns.
[0056] In the field of microfluidic chip technology, specific applications include, but are not limited to, the following products: High-efficiency micromixers: Microstructures that enhance fluid mixing efficiency are fabricated within microfluidic chips, achieved by processing a series of V-shaped channels in the mixing region. When fluid flows through these sloping structures, eddies and secondary flows are generated, breaking the laminar flow state and significantly improving mixing efficiency.
[0057] Cell sorting / enrichment device: This device utilizes microstructures to sort or enrich cells of different sizes. By fabricating wedge-shaped channels with specific tapers and widths, fluid dynamics effects are used to subject cells of different sizes to different forces as they flow through, thereby achieving separation.
[0058] In the field of micro-optical element technology, specific applications include, but are not limited to, the following products: Microlens arrays: Fabrication of microlens arrays for beam shaping, uniform illumination, and 3D sensing. A circular pattern is placed at the edge of the scanning field, and through multiple low-power scans, a smooth array of spherical or aspherical pits is created, with each pit serving as a microlens.
[0059] Diffractive optical elements: Manufacturing diffractive optical elements such as gratings. A grating structure is formed by processing a series of periodic V-shaped grooves to achieve modulation of light waves.
[0060] In the field of functional surface technology, specific applications include, but are not limited to, the following products: Biomimetic superhydrophobic surfaces: These are superhydrophobic surfaces that mimic the lotus leaf effect, used for self-cleaning, anti-icing, and drag reduction. By densely fabricating cone or wedge arrays on the material surface, these sharp microstructures effectively trap air, forming an air film that prevents water droplets from wetting the surface.
[0061] Structural color surfaces: Color is generated through the interference and diffraction of light by surface microstructures, used for anti-counterfeiting and decoration. Microgroove arrays with specific spacing and angles can produce vibrant and fade-resistant colors without the need for dyes.
[0062] Friction control surfaces: Surfaces with anisotropic frictional forces are manufactured for use in microrobots, precision jigs, etc. By machining an array of micro-wedge structures with specific orientations, the surface can be made to slide easily in one direction and difficult to slide in another.
[0063] Example 2 To create more complex or intricate structures (composite structures), this can be achieved through the combination and iteration of multiple static patterns. The difference between this embodiment and Embodiment 1 lies in the addition of the following step between steps S4 and S5: A smaller two-dimensional static pattern is superimposed on the formed three-dimensional structure, as referenced. Figure 4-5 For example, a basic square pyramid is first created using a larger square pattern, and then a smaller circular pattern is superimposed on its center and sharpened in step S5 to obtain a composite structure with a sharper bottom.
[0064] Therefore, the present invention adopts the above-mentioned AI-enabled laser tilting engraving method for preparing complex three-dimensional microstructures and its application. By controlling the position of the scanning pattern in the scanning field, the inherent beam tilting and converging effects are actively utilized to control the morphology of the three-dimensional structure.
[0065] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.
Claims
1. A method for fabricating complex three-dimensional microstructures using AI-enabled laser tilting engraving, characterized in that: Includes the following steps: S1. Establish a process database; S2. Input the data from the process database into the forward prediction model for training to improve the accuracy and efficiency of the forward prediction model; S3. Based on the requirements of the three-dimensional structure, use the forward prediction model to reverse-engineer the marking parameters of the two-dimensional static pattern. S4. Position the workpiece to be processed and set the two-dimensional static pattern at the target position; S5. Perform low-power multiple scanning processing on the two-dimensional static pattern to obtain the initial three-dimensional structure; S6. Iteratively optimize the three-dimensional structure initially formed in step S5.
2. The method for fabricating complex three-dimensional microstructures by AI-enabled laser tilting engraving according to claim 1, characterized in that: Step S1 specifically includes: S11. In the laser processing software, create a series of simple static test patterns and place the static test patterns at different coordinate positions within the scanning field; the different coordinate positions cover various areas from the center of the scanning field to the edge of the scanning field. S12. Set a set of fixed laser parameters and scan the static test pattern at each coordinate position in step S11 N times. S13. Using a surface profilometer or confocal microscope, accurately measure the three-dimensional morphological features of the grooves processed at each location; the three-dimensional morphological features are used to describe the complete shape information of the micro / nano structure in three-dimensional space, including the three-dimensional contour, shape changes and spatial positional relationships of the entire structure; S14. Associate the three-dimensional morphological features with the field coordinates of each static test pattern to form a mapping table or relationship curve, and construct the corresponding process database.
3. The method for preparing complex three-dimensional microstructures by AI-enabled laser tilting engraving according to claim 2, characterized in that: Step S2 specifically includes: S21. Perform data cleaning and feature engineering on the data in the process database; S22. Construct and train the positive prediction model using a deep neural network, and perform multiple verifications; S23. The positive prediction model is frozen using a genetic algorithm, and the input training data is continuously adjusted through an optimization algorithm until the three-dimensional structure data output by the positive prediction model is close to the actual three-dimensional structure data.
4. The method for preparing complex three-dimensional microstructures by AI-enabled laser tilting engraving according to claim 1, characterized in that: Step S3 specifically includes: S31. The user determines the three-dimensional structural data of the target three-dimensional structure according to the needs of the final product. S32. Based on the target three-dimensional structure, design a corresponding two-dimensional static pattern; S33. Based on the trained positive prediction model, obtain the optimal marking parameters corresponding to the target three-dimensional structure.
5. The method for preparing complex three-dimensional microstructures by AI-enabled laser tilting engraving according to claim 4, characterized in that: Step S33 includes: randomly generating a set of marking parameters; inputting the marking parameters into a forward prediction model to predict the three-dimensional structure data; when the predicted three-dimensional structure data does not match the three-dimensional structure data of the target three-dimensional structure, adjusting the marking parameter values using an optimization algorithm; repeating the iteration until the optimal marking parameters are found; and inputting the obtained optimal marking parameters into the forward prediction model for verification.
6. The method for preparing complex three-dimensional microstructures by AI-enabled laser tilting engraving according to claim 5, characterized in that: Step S4 specifically includes: S41. Fix the workpiece to be processed flat on the laser processing platform; S42. In the laser control software, drag and drop or position the two-dimensional static pattern to the target position at the edge of the scanning field obtained in step S33.
7. The method for preparing complex three-dimensional microstructures by AI-enabled laser tilting engraving according to claim 6, characterized in that: Step S5 specifically includes: S51. Set the laser power to a relatively low level and set the number of scans. The relatively low level is such that excessive energy in a single scan can prevent the material from undergoing severe vaporization and thermal damage. S52. Control the laser beam to repeatedly scan the two-dimensional static pattern that has been positioned at the target location; S53. Under the low-power multi-scan process, excess material is removed in a small amount and gradually to obtain the formed three-dimensional structure.
8. The method for preparing complex three-dimensional microstructures by AI-enabled laser tilting engraving according to claim 7, characterized in that: In step S53, the two-dimensional static pattern located at the edge of the scanning field, since each point of the two-dimensional static pattern is irradiated by the tilted beam, the common area of the tilted beam inside the material naturally forms a tilted, converging three-dimensional structure.
9. The method for preparing complex three-dimensional microstructures by AI-enabled laser tilting engraving according to claim 8, characterized in that: Step S6 specifically includes: S61. After completing the scanning process, replace the scanning pattern with a smaller and finer outline in the laser control software, and align it precisely with the previously processed target position. S62. The scanning pattern is scanned a small number of times with the same or lower power to obtain the final shaped three-dimensional structure.
10. The application of the laser-engraved micro / nano structures obtained by the AI-enabled laser tilting engraving method for fabricating complex three-dimensional microstructures as described in any one of claims 1-9 in the fields of biomedical devices, microfluidic chips, micro-optical elements, and functional surface technologies.