Preparation method and application of high-silicon 316L / Y2O3 medical composite material

Abstract

The application relates to the fields of biomedical metal materials and additive manufacturing technology, and more particularly to a preparation method and application of a high-silicon 316L / Y2O3 medical composite material. The preparation method comprises the following steps: mixing high-silicon 316L stainless steel powder and Y2O3 nanoparticles, wherein the silicon content of the high-silicon 316L stainless steel powder is 1.5-2.5 wt%, and the content of the Y2O3 nanoparticles is 0.3-1.0 wt%; an environmental agent model is constructed by using an XGBoost regression model, input feature vectors include laser power P, scanning speed v and physical descriptors derived based on a solidification heat transfer physical model; a physical consistency data enhancement strategy is used to expand the training samples; a Q-learning reinforcement learning algorithm is used to perform strategy optimization by taking relative density and microhardness as a double-target reward function, and optimal laser power P and scanning speed v are obtained; and the high-silicon 316L / Y2O3 medical composite material is prepared by a laser powder bed fusion forming process by using the optimal laser power P and scanning speed v obtained in the second step.

Description

Technical Field

[0001] This invention relates to the field of biomedical metallic materials and additive manufacturing technology, and more specifically to a method for preparing a high-silicon 316L / Y2O3 medical composite material and its application. Background Technology

[0002] In existing technologies, two main strategies are employed to improve the wear resistance of 316L stainless steel orthopedic implants: one is to promote the formation of a self-healing friction oxide film rich in SiO2 during sliding through alloying (such as increasing silicon content); the other is to add ceramic nanoparticles (such as Y2O3, TiC, TiB2, etc.) as hard reinforcing phases to improve surface hardness and wear resistance through dispersion strengthening. Laser powder bed fusion (LPBF) technology is widely used to form such metal matrix composites, and its process parameters (laser power P, scanning speed v, etc.) significantly affect density, microstructure, and mechanical properties.

[0003] The limitations of existing technologies include: the vast parameter space of LPBF processes (P, v, scanning spacing, layer thickness, etc.), making it difficult to efficiently find the globally optimal parameter combination that simultaneously optimizes density and hardness using traditional trial-and-error methods or orthogonal experiments, especially for composite materials with added nanoparticles where the parameter window is even narrower. In existing research, ceramic reinforcing phases (such as Y₂O₃) are typically considered passive hard phases, with wear resistance improved only through dispersion strengthening or dislocation pinning. However, there is a lack of understanding of their potential tribochemical reactions during friction (such as transformation into the Y-Si-O phase) and their active role in regulating the evolution of the wear subsurface microstructure. Traditional machine learning optimization methods are often "black box" models, lacking physical interpretability and prone to overfitting on small-scale experimental datasets. Summary of the Invention

[0004] This invention provides a method for preparing a high-silicon 316L / Y2O3 medical composite material and its application in enhancing wear resistance through friction-induced Y2SiO5. The aim is to provide a physically interpretable reinforcement learning (RL) framework to efficiently optimize the LPBF process of the high-silicon 316L / Y2O3 composite material, obtaining samples with high density and high hardness. Furthermore, it reveals a novel wear-resistant mechanism in which the composite material, during sliding wear, generates a Y2O3 / Y2SiO5 composite structure in situ through friction-induced friction, thereby stabilizing the wear subsurface and inhibiting microcrack initiation. This provides a new material with both excellent wear resistance and biocompatibility for biomedical implants.

[0005] The above objectives are achieved through the following technical solutions: A method for preparing a high-silicon 316L / Y2O3 medical composite material includes the following steps: Step 1, Composite Powder Preparation: High-silicon 316L stainless steel powder is mixed with Y2O3 nanoparticles to obtain composite powder, wherein the silicon content of the high-silicon 316L stainless steel powder is 1.5-2.5 wt%, and the content of the Y2O3 nanoparticles is 0.3-1.0 wt%; the matrix Si content is ≥1.5 wt% to ensure sufficient Si diffusion to the Y2O3 particle interface during friction; if the Y2O3 content is less than 0.3 wt%, the number of Y2SiO5 particles will be insufficient, and if it is higher than 1.0 wt%, agglomeration is likely to occur; Step 2, Process Parameter Optimization: An environmental proxy model is constructed using the XGBoost regression model. The input feature vector includes laser power P, scanning speed v, and physical descriptors derived from the solidification heat transfer physics model. The physical descriptors include volumetric energy density E, temperature gradient G, and cooling rate dT / dt. The training samples are expanded using a physical consistency data augmentation strategy. The Q-learning reinforcement learning algorithm is used to optimize the strategy using relative density and microhardness as dual-objective reward functions to obtain the optimal laser power P and scanning speed v. Step 3, LPBF forming: Using the optimal laser power P and scanning speed v obtained in step S2, the high silicon 316L / Y2O3 medical composite material is prepared by laser powder bed melting forming process.

[0006] The mixing process employs ultrasonic vibration mixing, with the following ultrasonic vibration parameters: ultrasonic frequency 40-60kHz, ultrasonic power 200-400W, and ultrasonic time 30-90min.

[0007] Physical consistency data enhancement includes: applying a Gaussian perturbation with a standard deviation σ = 0.05 to the laser power P and scanning speed v, and synchronously updating the physical descriptor according to the physical formula.

[0008] The formula for calculating volumetric energy density E is: E=P / (v×h×t) Where h is the scanning interval and t is the layer thickness.

[0009] The formula for calculating the temperature gradient G is: G = η × P / (2π × k × r × v) Where η is the laser absorptivity, ranging from 0.6 to 0.7; k is the thermal conductivity; and r is the characteristic radius of the molten pool.

[0010] The Q-learning algorithm has a learning rate α=0.1, a discount factor γ=0.9, and an exploration rate ε=0.2.

[0011] The optimal laser power P ranges from 160 to 240 W, and the scanning speed v ranges from 1000 to 1500 mm / s.

[0012] The relative density of the composite material is ≥99%, and its microhardness is HV. 0.5 ≥280.

[0013] During the sliding friction process, Y2O3 / Y2SiO5 composite particles are generated in situ on the subsurface layer of the worn surface of the above composite material.

[0014] The aforementioned composite material is used in orthopedic implants, which are selected from bone screws, bone plates, hip joint prostheses, or knee joint prostheses.

[0015] The beneficial effects of the preparation method and application of the high-silicon 316L / Y2O3 medical composite material of the present invention are as follows: 1. High-efficiency intelligent process optimization: For the first time, physically interpretable reinforcement learning is applied to LPBF forming of high-silicon 316L / Y2O3 composite materials. It can autonomously find the global optimal process window (P=200W, v=1343mm / s) with only a small amount of initial experimental data (49 sets), while maximizing the relative density and microhardness, which is better than traditional trial and error methods and local optima.

[0016] 2. Significantly improved mechanical properties: The grain size of the RL-optimized sample is refined, resulting in a substantial increase in microhardness compared to conventional LPBF316L. The density of geometrically necessary dislocations is higher, and the grain boundary strengthening effect is significant.

[0017] 3. Excellent wear resistance: Under both high-load low-speed and low-load high-speed conditions, the RL-optimized sample exhibits the lowest coefficient of friction (approximately 0.46 under high-load low-speed conditions, compared to approximately 0.68 for pure 316L), with minimal fluctuation. Wear weight loss is reduced by 58%–67% compared to pure 316L (depending on operating conditions), with the shallowest wear track depth and the least ploughing and spalling.

[0018] 4. A novel wear-resistant mechanism was revealed: a dynamic formation mechanism of friction-induced in-situ Y₂O₃ / Y₂SiO₅ composite structure was discovered. These composite particles not only act as a hard phase themselves, but more importantly, they strongly pin dislocations, inhibiting the expansion of plastic shear bands; they confine stress-induced α'-martensite phase transformation to a narrow region between particles, avoiding volume expansion and microcrack initiation caused by large-scale phase transformation, and significantly improving the damage tolerance of the wear subsurface. This mechanism breaks through the traditional view of a "passive hard phase" and provides new insights for designing highly wear-resistant metal matrix composites.

[0019] 5. Biomedical prospects: The composite material obtained by this invention maintains the austenitic structure, combined with the good biocompatibility and corrosion resistance brought by the high silicon content (based on 316L), and avoids the introduction of toxic or allergenic elements through RL optimization, making it particularly suitable for long-term implants such as orthopedic internal fixation devices (such as bone screws and bone plates). Attached Figure Description

[0020] Figure 1 This is a schematic diagram illustrating the optimization of LPBF process parameters using the RL framework; Figure 2 It shows the morphology, particle size distribution, and EDS elemental distribution of the composite powder. Figure 3 This is a graph showing the prediction performance of the XGBoost model and the importance of SHAP features; Figure 4 This is a graph showing the convergence and Pareto front of Q-learning training; Figure 5 These are SEM / EBSD tissue comparison images under different parameters; Figure 6 It is a comparison chart of process parameters, microstructure, and hardness; Figure 7 This is a TEM characterization (including SAED) image of the RL-optimized sample; Figure 8 This is a comparison chart of the evolution of friction coefficient and wear amount; Figure 9 It is a three-dimensional wear track morphology and quantitative wear analysis diagram; Figure 10 These are SEM and EDS elemental distribution maps of the worn surface; Figure 11 These are comparison images of surface morphology under high load and low speed wear. Figure 12 It is a microstructure characterization diagram of the cross-section of the friction interface; Figure 13 These are subsurface TEM images and α'-martensite structure diagrams; Figure 14 This is a diagram showing the interaction between Y-Si-O particles and dislocations; Figure 15 This is a schematic diagram of the collaborative wear resistance mechanism. Detailed Implementation

[0021] A method for preparing a high-silicon 316L / Y2O3 medical composite material includes six steps: composite powder preparation, physically interpretable reinforcement learning process optimization, LPBF forming, microstructure characterization, tribological testing, and wear resistance mechanism analysis.

[0022] Step 1: Preparation of composite powder (1) The matrix powder is high silicon 316L stainless steel powder with a Si content of 2wt% and the remaining components conform to the 316L stainless steel standard. The powder morphology is gas atomized spherical and the particle size distribution is suitable for laser powder bed melting and forming.

[0023] (2) The reinforcing phase is Y2O3 nanoparticles with a purity of ≥99.9% and an average particle size of 120±5nm.

[0024] (3) The mixture was mixed by ultrasonic vibration at a frequency of 50 kHz and a power of 300 W for 1 h to obtain a composite powder of 0.5 wt% Y2O3 and 99.5 wt% high silicon 316L.

[0025] like Figure 2 As shown, (a) morphology of high-silicon 316L / 0.5 wt% Y2O3 composite powder; (b) particle size distribution; (c, d) EDS elemental distribution maps (Si, Ni, Cr, Fe); (e) elemental composition of high-silicon 316L alloy determined by ICP analysis.

[0026] Step 2: Construction of a Physically Interpretable Reinforcement Learning Framework and Optimization of LPBF Process (1) Define the state space S: Discretize the laser power P and the scanning speed v to form a grid state; wherein the laser power P ranges from 140 to 260W with a step size of 20W; the scanning speed v ranges from 700 to 1600mm / s with a step size of 150mm / s.

[0027] (2) Define the action space A: make incremental adjustments to the laser power P and scanning speed v in 8 directions in the current state.

[0028] (3) Constructing an environmental proxy model: The XGBoost regression model was trained using 49 sets of experimental data; the experimental data included different combinations of P and v and their corresponding relative density ρ and microhardness HV. 0.5 In addition to laser power P and scanning speed v, the input features include physical descriptors such as volumetric energy density E, temperature gradient G, and cooling rate dT / dt. The volumetric energy density E is calculated as E=P / (v·h·t), where h is the scanning distance and t is the layer thickness. The temperature gradient G and cooling rate dT / dt are calculated based on the Rosenthal analytical model.

[0029] (4) Data augmentation: Gaussian perturbation is applied to the laser power P and scanning speed v, with a standard deviation σ=0.05. The volume energy density E, temperature gradient G and cooling rate dT / dt are recalculated synchronously according to the physical formula to ensure that the augmented data conforms to the law of energy conservation and solidification.

[0030] (5) Reward function design: The reward function R is composed of normalized relative density and microhardness weighted, and a dual-objective reward strategy is adopted.

[0031] (6) Reinforcement learning optimization: The Q-learning algorithm is used to iteratively update the Q table through the Bellman equation until the Q table converges, and the global optimal process parameters corresponding to the maximum Q value are obtained.

[0032] like Figure 1 The diagram illustrates a physically interpretable reinforcement learning framework for optimizing laser powder bed melting process parameters. The RL agent autonomously explores the parameter spaces of laser power (P) and scanning velocity (v). An XGBoost surrogate model, trained on experimental data and enhanced with physical descriptors (such as volumetric energy density E, temperature gradient G, and cooling rate dT / dt), acts as the environment to predict material properties (relative density ρ and microhardness HV0.5) and provides a dual-objective reward signal. This feedback guides the agent to converge to the optimal process window that simultaneously maximizes density and hardness.

[0033] like Figure 3 As shown, this is the prediction performance of the optimized XGBoost model for relative density and microhardness. (a, b) are scatter plots of predicted values ​​versus experimental values ​​on the training and test sets, respectively; (c, d) are bar charts showing the importance of SHAP features for density and hardness predictions, respectively, quantitatively displaying the average contribution of each input feature.

[0034] like Figure 4 The diagram shows the Q-learning training process and optimization results. (a) shows the convergence curve of reward with the number of training rounds; (b) shows the convergence behavior under different state space resolutions; (c) shows a schematic diagram of the dual-objective reward signal composition; (d) shows a heatmap of the maximum Q value (Q_max) on the entire Pv plane, indicating that the high Q region is concentrated in a narrow process window, and the global maximum value is located near P=200 W and v=1343 mm / s; (e) shows a comparison of the Pareto front, where the RL optimized parameters (marked with red solid marks) surpass all traditional experimental points and the original Pareto front in both density and hardness.

[0035] Step 3: LPBF forming (1) Sample preparation was carried out using the global optimal parameters recommended by reinforcement learning in step two; the optimal laser power P was 200W and the scanning speed v was 1343mm / s.

[0036] (2) Fixed process parameters: layer thickness is 30μm, scanning spacing is 0.08mm, substrate temperature is 80℃, and a 67° interlayer rotation scanning strategy is adopted.

[0037] (3) Sample preparation: The sample size is 5mm×5mm×5mm.

[0038] (4) In contrast, local optimal parameters or pure high-silicon 316L powder are used for forming.

[0039] like Figure 5The figure shows the SEM microstructure and EBSD-IPF (inverse pole figure) of the high-silicon 316L / Y2O3 composite material under different process parameters. (ac) represents the RL optimization parameters; (df) represents local optimum 1; (gi) represents local optimum 2.

[0040] like Figure 6 As shown, the key parameters, microstructure characteristics, and hardness of the materials under different processing conditions are compared. RL optimization parameters yielded the minimum grain size and the highest hardness, while the Y2O3 reinforcing phase further improved the hardness.

[0041] Step 4: Microstructural Characterization (1) The morphology of the grains was observed using a scanning electron microscope (SEM).

[0042] (2) Electron backscatter diffraction (EBSD) was used to analyze grain size, dislocation density and martensitic phase transformation distribution.

[0043] (3) High-resolution structures were observed using transmission electron microscopy (TEM), including high-resolution TEM (HRTEM) and selected area electron diffraction (SAED), to analyze the distribution of Y-Si-O nanoparticles.

[0044] (4) The density of geometrically necessary dislocations is analyzed using the nuclear average orientation difference (KAM) diagram.

[0045] like Figure 7 The image shows the transmission electron microscopy characterization of the RL-optimized sample. (a) is a bright-field image showing the submicron-scale cellular substructure; (b) is a higher-magnification image showing the intracellular dislocation network and stacking faults; (c) is a high-resolution TEM image clearly showing the stacking faults in the face-centered cubic (FCC) austenitic matrix; (d) is the corresponding selected area electron diffraction pattern, labeled as an FCC structure.

[0046] Step 5: Friction and Wear Test (1) A ball-disc friction and wear tester was used, and the grinding ball material was 316L stainless steel ball.

[0047] (2) Two service conditions were set up for comparative testing: the high load and low speed conditions were 15N load, 1.5mm wear radius, 5.28m / min sliding speed, and 10min test time; the low load and high speed conditions were 5N load, 4mm wear radius, 14.07m / min sliding speed, and 10min test time.

[0048] (3) Record the coefficient of friction (COF) in real time.

[0049] (4) The following analyses were performed after the test: the wear loss weight was measured; the three-dimensional wear morphology was obtained by laser confocal microscopy; the wear cross-sectional depth was measured; and the wear surface and sub-section were analyzed by SEM / EDS / EBSD / TEM.

[0050] like Figure 8 The figure shows the evolution of friction coefficient and volumetric wear of the RL-optimized, locally optimal, and pure 316L samples under two wear conditions. (a, c) represent high load and low speed conditions (15 N, 5.28 m / min); (b, d) represent low load and high speed conditions (5 N, 14.07 m / min). The RL-optimized sample consistently exhibits the lowest friction coefficient, the smallest fluctuation, and the lowest wear.

[0051] like Figure 9 The figures show the three-dimensional wear track morphology and quantitative wear analysis of the RL-optimized, locally optimal, and pure 316L samples. (ac) represents the three-dimensional surface morphology under low-load, high-speed conditions; (df) represents the three-dimensional surface morphology under high-load, low-speed conditions; (g) represents the wear track cross-sectional depth profile extracted from the high-speed-low-stress condition; and (hi) represents the wear weight loss statistics under high-speed-low-stress and high-stress-low-speed conditions, respectively. The RL-optimized sample exhibits the lowest wear weight loss.

[0052] like Figure 10 The image shows the scanning electron microscope (SEM) morphology and energy dispersive spectroscopy (EDS) elemental distribution of the sample surfaces after low-load, high-speed wear. All samples developed silicon-rich (Si) and oxygen-rich (O) oxides on their worn surfaces. The oxide layer of the RL-optimized sample was more tightly bonded to the substrate, with no significant peeling.

[0053] like Figure 11 The image shows the wear surface morphology under high load and low speed sliding conditions. (ac) represents RL-optimized high-silicon 316L / Y2O3; (df) represents locally optimal high-silicon 316L / Y2O3; and (gi) represents pure high-silicon 316L. The wear damage shows an increasing progression: the RL-optimized sample has only slight scratches and a small amount of oxide debris; the locally optimal sample shows ploughing grooves and spalling; and the pure 316L sample exhibits wide and deep ploughing grooves, plastic accumulation, and dense microcracks.

[0054] Step Six: Wear Resistance Mechanism Analysis (1) The wear subsurface section sample was prepared by focused ion beam (FIB).

[0055] (2) Transmission electron microscopy (TEM), selected area electron diffraction (SAED) and energy dispersive spectroscopy (EDS) were used to identify friction-induced Y2O3 / Y2SiO5 composite particles.

[0056] (3) Observe the pinning effect of the Y2O3 / Y2SiO5 composite particles on dislocations.

[0057] (4) Analyze the spatial confinement effect of the composite particles on the α'-martensite phase transformation.

[0058] like Figure 12 The figures show the microstructure characterization of the cross-section of the friction interface. (a) SEM image and EDS elemental distribution of Cr, Fe, Si, Y, and Mo; (b) EBSD phase diagram showing the distribution of FCC austenite (red), BCC martensite (green), SiO2 (blue), and Y2O3 (yellow); (c) nuclear average orientation difference (KAM) diagram revealing the lattice distortion within the friction layer; and (d) inverse pole figure (IPF) diagram.

[0059] like Figure 13 As shown, the subsurface microstructure beneath the wear track of RL-optimized high-silicon 316L / Y2O3 is presented. (a) is a cross-sectional TEM image and the corresponding EDS elemental distribution, showing the distribution of Y-Si-O particles in the untransformed FCC matrix; (bc) is a high-magnification TEM image of nanolamellar α'-martensite with a lamellar spacing of tens of nanometers; (d) is a selected area electron diffraction pattern, confirming the BCC crystal structure.

[0060] like Figure 14 As shown, wear affects the interaction between Y-Si-O composite particles and dislocations in the subsurface. (a) is the TEM view of the embedded oxide particles; (b) is a high-magnification image showing that the dislocation lines terminate at the particle-matrix interface, indicating a strong pinning effect; (cd) is the selected area electron diffraction pattern, indicating the coexistence of Y2O3 and Y2SiO5 phases.

[0061] like Figure 15 The diagram shows a schematic of the synergistic wear resistance mechanism. During sliding wear, the RL-optimized composite material forms Y2O3 / Y2SiO5 composite particles in situ through tribochemical reactions. These particles pin dislocations and restrict the α'-martensite phase transformation in local areas, thereby stabilizing the wear subsurface and inhibiting the initiation of microcracks, achieving a balance between high hardness and excellent damage tolerance.

[0062] Example 1, specific parameters of the above preparation method: Step 1: Preparation of composite powder Take 99.5g of high-silicon 316L powder (Si content 2wt%, particle size 15-45μm) and 0.5g of Y2O3 nanopowder (average particle size 120nm). Place the two in an ultrasonic mixer and mix for 1 hour at an ultrasonic frequency of 50kHz and a power of 300W to obtain 0.5wt% Y2O3 / high-silicon 316L composite powder.

[0063] Step 2: Reinforcement learning to optimize process parameter acquisition Forty-nine mesh experiments were conducted beforehand, with laser power P ranging from 140 to 260 W and scanning speed v ranging from 700 to 1600 mm / s. The relative density and microhardness HV corresponding to each set of process parameters were measured. 0.5 .

[0064] The XGBoost model is trained using P, v, volumetric energy density E, temperature gradient G, and cooling rate dT / dt as input features.

[0065] Construct a Q-learning environment, setting the state discretization granularity, action set (ΔP=±10, ±20W; Δv=±50, ±100mm / s), and reward function R=ρ_norm+HV_norm.

[0066] Training continues until the Q-table converges, and the process parameters corresponding to the global maximum Q-value are extracted: laser power P = 200W, scanning speed v = 1343mm / s.

[0067] Step 3: LPBF forming On the LPBF device, a 5×5×5mm cubic sample was formed using a scanning strategy with a layer thickness of 30μm, a scanning spacing of 0.08mm, a substrate preheating temperature of 80℃, and an interlayer rotation of 67°, according to reinforcement learning optimization parameters (laser power 200W, scanning speed 1343mm / s), as well as the sample required for subsequent wear testing.

[0068] Step 4: Post-processing The sample was removed by wire cutting, treated with glass bead sandblasting, and ultrasonically cleaned with ethanol.

[0069] Step 5: Performance Testing The relative density was measured using Archimedes' displacement method, and the result was greater than 99.2%.

[0070] Microhardness was measured using a Vickers hardness tester with a load of 0.5 kgf, and the result was HV. 0.5 Greater than 280.

[0071] The wear resistance was tested using a ball-disc wear test (load 15N, sliding speed 5.28m / min). The results showed that the coefficient of friction was stable at around 0.46, and the wear weight loss was reduced by about 58% compared with pure 316L.

[0072] Step 6: Microstructure Analysis EBSD analysis showed that the average grain size was about 8 μm and the KAM value was relatively high.

[0073] TEM observation of the wear subsurface layer revealed an α'-martensite layer with a thickness of approximately 1-2 μm, in which Y2O3 / Y2SiO5 particles with a diameter of approximately 100-500 nm are dispersed, and dislocation lines terminate at the particle interface.

[0074] SAED analysis confirmed the presence of BCC martensite and Y-Si-O phase.

[0075] Comparative example: Using locally optimal parameters (such as laser power P=220W, scanning speed v=1100mm / s) or pure high-silicon 316L powder (without Y2O3) and forming according to the same LPBF process, the comparison results show that the samples prepared using locally optimal parameters or without Y2O3 powder have significantly lower density, hardness and wear resistance than the reinforcement learning optimized samples.

[0076] .

Claims

1. A method for preparing a high-silicon 316L / Y2O3 medical composite material, characterized in that, Includes the following steps: Step 1, Composite Powder Preparation: High-silicon 316L stainless steel powder is mixed with Y2O3 nanoparticles to obtain composite powder, wherein the silicon content of the high-silicon 316L stainless steel powder is 1.5-2.5 wt%, and the content of the Y2O3 nanoparticles is 0.3-1.0 wt%. Step 2, Process Parameter Optimization: An environmental proxy model is constructed using the XGBoost regression model. The input feature vector includes laser power P, scanning speed v, and physical descriptors derived from the solidification heat transfer physics model. The physical descriptors include volumetric energy density E, temperature gradient G, and cooling rate dT / dt. The training samples are expanded using a physical consistency data augmentation strategy. The Q-learning reinforcement learning algorithm is used to optimize the strategy using relative density and microhardness as dual-objective reward functions to obtain the optimal laser power P and scanning speed v. Step 3, LPBF forming: Using the optimal laser power P and scanning speed v obtained in Step 2, the high-silicon 316L / Y2O3 medical composite material is prepared by laser powder bed melting forming process.

2. The preparation method according to claim 1, characterized in that, In step one, the mixing is performed using ultrasonic vibration. The ultrasonic vibration process parameters are: ultrasonic frequency 40-60kHz, ultrasonic power 200-400W, and ultrasonic time 30-90min.

3. The preparation method according to claim 1, characterized in that, The physical consistency data enhancement in step two includes: applying a Gaussian perturbation with a standard deviation σ = 0.05 to the laser power P and the scanning speed v, and synchronously updating the physical descriptor according to the physical formula.

4. The preparation method according to claim 1, characterized in that, The formula for calculating the volumetric energy density E in step two is as follows: E=P / (v×h×t) Where h is the scanning interval and t is the layer thickness.

5. The preparation method according to claim 1, characterized in that, The formula for calculating the temperature gradient G in step two is: G = η × P / (2π × k × r × v); Where η is the laser absorptivity, ranging from 0.6 to 0.7; k is the thermal conductivity; and r is the characteristic radius of the molten pool.

6. The preparation method according to claim 1, characterized in that, The Q-learning algorithm described in step two has a learning rate α=0.1, a discount factor γ=0.9, and an exploration rate ε=0.

2.

7. The preparation method according to claim 1, characterized in that, The optimal laser power P mentioned in step three is in the range of 160-240W, and the scanning speed v is in the range of 1000-1500mm / s.

8. The high-silicon 316L / Y2O3 medical composite material prepared by the preparation method according to any one of claims 1-7, characterized in that, The high-silicon 316L / Y2O3 medical composite material has a relative density ≥99% and a microhardness HV. 0.5 ≥280.

9. The composite material according to claim 8, characterized in that, During the sliding friction process, Y2O3 / Y2SiO5 composite particles are generated in situ on the subsurface layer of the worn surface of the high-silicon 316L / Y2O3 medical composite material.

10. The composite material according to claim 9, characterized in that, Used in orthopedic implants, the orthopedic implants being selected from bone screws, bone plates, hip joint prostheses, or knee joint prostheses.

Images