Method and system for simulating and optimizing performance of medical biomaterials based on digital twinning

Abstract

This invention discloses a method and system for simulating and optimizing the performance of medical biomaterials based on digital twins, belonging to the field of medical bio-implant material simulation technology. First, the physical morphological characteristics and biochemical environmental characteristics of the target implant are acquired to form a unique multidimensional fusion feature set representing the combined relationship between the implant and its environment. An initial digital twin model is then constructed based on this set, and dynamic simulation of the material evolution process within a preset service life is performed to obtain the mass degradation sequence and stress fatigue sequence. Subsequently, a sequence analysis model is used to extract multidimensional temporal evolution laws and determine the dynamic performance change characteristics. When the performance does not reach the safe service performance threshold, a parameter optimization algorithm is used to iteratively optimize within the candidate process parameter space to obtain the target material ratio and molding process. Based on this, the twin model is updated to output the full life-cycle service state prediction results, significantly improving the accuracy of implant degradation trend prediction and the targeted nature of process optimization.

Description

Technical Field

[0001] This invention relates to the field of medical bio-implant material simulation technology, and in particular discloses a method and system for simulating and optimizing the performance of medical bio-materials based on digital twins. Background Technology

[0002] Performance simulation and process optimization of medical bio-implant materials are crucial steps in biomaterials research and development, directly impacting the safety of implants in clinical service. A search revealed that existing patent applications CN115270558A and CN114564861A disclose biomaterial simulation and optimization technologies that primarily simulate material physical properties or static physiological environments. These technologies fail to achieve a fusion mapping between the implant's physical morphology and the characteristics of the human biochemical environment, and lack digital twin dynamic simulation, degradation-stress time-series feature extraction, and iterative optimization mechanisms for process parameters.

[0003] The shortcomings and consequences of existing technologies are as follows: First, the microstructure characteristics of materials are modeled separately from biochemical environmental parameters such as human body pH and body fluid corrosion, resulting in a large deviation between simulated working conditions and the actual implantation environment; second, it is impossible to simultaneously simulate the dynamic coupling process of material quality degradation and stress fatigue during the service life, making it difficult to extract the temporal evolution law; third, there is a lack of automatic iterative optimization methods for material ratio and molding process based on performance degradation characteristics, and research and development relies on a large number of physical experiments; fourth, a closed-loop update and optimization system for the model has not been built, resulting in insufficient accuracy in predicting the service status throughout the entire life cycle, leading to high research and development costs and long cycles, making it difficult to meet the clinical requirements for long-term safe service.

[0004] Therefore, how to integrate the microstructure parameters of materials with the working conditions such as the human body fluid environment and dynamic stress to achieve accurate simulation of the material's service state and automatically complete the iterative optimization of material ratio and process parameters has become a key issue to break through the high cost barrier of testing and meet the clinical long-term service requirements. Summary of the Invention

[0005] This invention provides a method and system for simulating and optimizing the performance of medical biomaterials based on digital twins, aiming to solve the shortcomings of the existing technology, such as the reliance on physical sample preparation for the research and development of medical implant materials, long cycle, high cost, inability to accurately reproduce the complex biochemical environment in vivo, difficulty in predicting the coupled failure of material degradation and stress fatigue, and reliance on experience to adjust process parameters.

[0006] One aspect of this invention relates to a method for simulating and optimizing the performance of medical biomaterials based on digital twins, comprising the following steps: S100. Obtain the set of physical morphological features corresponding to the target implant and the set of biochemical environmental features corresponding to the target implantation site; the set of physical morphological features includes pore distribution features, and the set of biochemical environmental features includes pH features. S200. A classification and regression algorithm is used to perform feature mapping and fusion processing on the physical morphology feature set and the biochemical environment feature set to obtain a multi-dimensional fused feature set; each multi-dimensional fused feature set corresponds to a unique combination relationship between the implant and the implantation environment. S300: An initial digital twin model is constructed based on a multi-dimensional fusion feature set. The initial digital twin model is used to simulate the material evolution process of the target implant in a real physical environment. S400: Dynamically simulate the service process of the target implant within the preset service period using an initial digital twin model to obtain the mass degradation sequence and stress fatigue sequence corresponding to the target implant. S500: Input the mass degradation sequence and stress fatigue sequence into the sequence analysis model to extract the multidimensional time-series evolution law and determine the dynamic performance change characteristics of the target implant; the dynamic performance change characteristics are used to characterize the degradation trend of the target implant material in complex environments; S600. If the dynamic performance change characteristics do not reach the preset safe service performance threshold, then the dynamic performance change characteristics are used as input parameters, and the parameter optimization algorithm is used to perform iterative optimization processing in the preset candidate process parameter space to obtain the target material ratio and the target molding process. S700 updates the initial digital twin model based on the target material ratio and target molding process to obtain the target digital twin model, and outputs the full life cycle service status prediction results of the target implant through the target digital twin model.

[0007] Further, step S100 includes: S110. Extract the set of physical morphological features corresponding to the target implant; S120. Based on the set of physical morphological features, obtain the set of biochemical environmental features corresponding to the target implantation site; S130. If the initial degradation rate of the material corresponding to the biochemical environmental feature set is greater than the preset degradation threshold, then collect the target ion concentration. S140. Calculate the fusion matching degree value based on the target ion concentration; S150. Based on the fusion matching degree value, obtain the set of physical morphological features corresponding to the target implant and the set of biochemical environmental features corresponding to the target implantation site.

[0008] Further, step S200 includes: S210. Extract the surface roughness feature values ​​contained in the physical morphology feature set and the local shear stress feature values ​​contained in the biochemical environment feature set; S220. Based on the surface roughness characteristic value and local shear stress, node division is performed to obtain the initial classification label of rejection risk. S230. If the initial classification label is a high-risk rejection label, collect the concentration of degradation products and the osmotic pressure of tissue fluid, and perform fitting calculation on the concentration of degradation products and the osmotic pressure of tissue fluid to obtain the feature mapping weight matrix. S240. The physical morphological feature set and the biochemical environment feature set are fused using a mapping weight matrix to obtain a multi-dimensional fused feature set.

[0009] Further, step S300 includes: S310. Analyze the multi-dimensional fusion feature set and extract spatial geometric features and material property features; S320. Generate a mesh topology based on spatial geometric features and establish the material constitutive equation corresponding to the target implant; S330. By combining the material constitutive equations and solving them, the spatial stress distribution is obtained. S340. If the spatial stress distribution exceeds the preset stress threshold, generate the predicted material evolution trajectory. S350. Based on the evolution trajectory prediction results and grid topology, an initial digital twin model is constructed.

[0010] Further, step S400 includes: S410. Apply a dynamic load spectrum to the initial digital twin model to perform simulation calculations and obtain time-series simulation state data; S420. Extract fluid shear force from time-series simulation state data and obtain mass loss distribution; S430. The stiffness attenuation field is obtained by solving based on the mass loss distribution, and the cyclic stress response is determined based on the stiffness attenuation field. S440. Calculate the fatigue damage value based on the cumulative cyclic stress response, record the mass loss distribution and fatigue damage value, and obtain the mass degradation sequence and stress fatigue sequence corresponding to the target implant.

[0011] Further, step S500 includes: S510. Obtain the mass degradation sequence and stress fatigue sequence, process the mass degradation sequence and stress fatigue sequence, and generate a multi-dimensional time series feature vector. S520. Input the multidimensional time series feature vector into the sequence analysis model, and extract the multidimensional time series evolution law through the sequence analysis model; S530. Determine the dynamic performance change characteristics of the target implant based on the multidimensional temporal evolution law; S540: Construct an exponential decay function based on dynamic performance change characteristics, and output the degradation trend of the target implant material in complex environments through the exponential decay function.

[0012] Further, step S600 includes: S610. If the dynamic performance change characteristics do not reach the preset safe service performance threshold, calculate the performance deviation value and generate a performance deviation vector. S620. Construct an optimization fitness function based on the performance deviation vector, and perform iterative optimization operations in the preset candidate process parameter space through the optimization fitness function to obtain the global optimal solution set; S630. Analyze the global optimal solution set and determine the target material ratio and target molding process corresponding to the global optimal solution set.

[0013] Further, step S700 includes: S710. Extract the material property matrix and process feature vector. The material property matrix and process feature vector are generated from the target material ratio and the target molding process. S720. Update the initial digital twin model based on the material property matrix and process feature vector to obtain the target digital twin model; S730: Input the service environment loads into the target digital twin model, perform simulation and deduction calculations, and generate a full life cycle state evolution sequence; S740. Dimensionality reduction is performed on the entire life cycle state evolution sequence to obtain fatigue damage characteristics; S750. If the fatigue damage characteristics meet the preset damage judgment conditions, the full life cycle service status prediction results of the target implant will be output.

[0014] Another aspect of the present invention relates to a digital twin-based system for simulating and optimizing the performance of medical biomaterials, used to implement the aforementioned digital twin-based method for simulating and optimizing the performance of medical biomaterials, comprising: The biochemical environment feature set acquisition module is used to acquire the physical morphological feature set corresponding to the target implant and the biochemical environment feature set corresponding to the target implantation site; the physical morphological feature set includes pore distribution features, and the biochemical environment feature set includes pH features. The multidimensional fusion feature set acquisition module is used to perform feature mapping and fusion processing on the physical morphology feature set and the biochemical environment feature set using a classification and regression algorithm to obtain a multidimensional fusion feature set; each multidimensional fusion feature set corresponds to a unique combination relationship between the implant and the implantation environment. The initial digital twin model construction module is used to construct an initial digital twin model based on a multi-dimensional fusion feature set. The initial digital twin model is used to simulate the material evolution process of the target implant in a real physical environment. The mass degradation sequence and stress fatigue sequence acquisition module is used to perform dynamic simulation of the service process of the target implant within a preset service period through an initial digital twin model, and to obtain the mass degradation sequence and stress fatigue sequence corresponding to the target implant. The dynamic performance change characteristic determination module is used to input the mass degradation sequence and stress fatigue sequence into the sequence analysis model, extract the multidimensional time-series evolution law, and determine the dynamic performance change characteristics of the target implant; the dynamic performance change characteristics are used to characterize the degradation trend of the target implant material in complex environments; The target material ratio and target molding process acquisition module is used to obtain the target material ratio and target molding process by using the dynamic performance change characteristics as input parameters and employing a parameter optimization algorithm to perform iterative optimization within the preset candidate process parameter space if the dynamic performance change characteristics do not reach the preset safe service performance threshold. The target digital twin model acquisition module is used to update the initial digital twin model based on the target material ratio and the target molding process to obtain the target digital twin model, and output the prediction results of the full life cycle service status of the target implant through the target digital twin model.

[0015] The beneficial effects achieved by this invention are as follows: The present invention provides a method and system for simulating and optimizing the performance of medical biomaterials based on digital twins, which aims to solve the problem in traditional implant design that it is difficult to accurately predict the long-term degradation trend of materials in complex biochemical environments and optimize the material ratio and molding process accordingly. This invention first obtains the physical morphological characteristics of the target implant, such as pore distribution, and the biochemical environmental characteristics of the implantation site, such as pH. A classification and regression algorithm is then used to map and fuse these two types of features, forming a unique multidimensional fused feature set characterizing the relationship between the implant and its environment. Based on this, an initial digital twin model is constructed to dynamically simulate the material evolution process within a preset service life, obtaining the mass degradation sequence and stress fatigue sequence. Subsequently, a sequence analysis model is used to extract multidimensional temporal evolution patterns and determine dynamic performance change characteristics. When these characteristics do not reach the safe service performance threshold, a parameter optimization algorithm iteratively optimizes within the candidate process parameter space to obtain the target material ratio and molding process. Based on this, the twin model is updated to output the full life-cycle service status prediction results, thereby significantly improving the accuracy of implant degradation trend prediction and the targeting of process optimization. This provides a scientific basis for the safe and reliable service of implants. The specific beneficial effects achieved are as follows: 1. Significantly improve the accuracy and reliability of simulation and evaluation of the service performance of biomaterials. This invention simultaneously collects physical morphological features such as the pore distribution of implants and biochemical environmental features such as the pH of the implantation site, achieving comprehensive data characterization of the material's intrinsic properties and the complex in vivo service environment. This overcomes the shortcomings of traditional simulation modeling, which only considers single structural parameters and ignores the coupling effects of the biochemical environment, providing a complete data foundation for high-precision modeling. Through classification and regression algorithms, it achieves precise mapping and fusion of physical and biochemical features, constructing a multi-dimensional fusion feature set that uniquely corresponds to the combination of implant and implantation environment. This solves the problems of single-dimensional feature processing, ambiguous coupling relationships, and poor adaptability, ensuring the uniqueness and accuracy of the model's input features. Based on the fusion features, an initial digital twin model is constructed, realistically replicating the material evolution process under physical conditions and improving the model's fit with actual service scenarios. Through dynamic simulation throughout the entire service cycle, it continuously outputs quality degradation and stress fatigue sequences, achieving refined and temporal capture of the material degradation process, avoiding the shortcomings of traditional static evaluation that cannot reflect dynamic evolution. Relying on sequence analysis models to extract multi-dimensional temporal evolution patterns, it accurately characterizes the dynamic degradation trend of materials under complex environments, effectively improving the dynamism and accuracy of performance evaluation. By accurately identifying substandard operating conditions through threshold determination, and relying on iterative optimization algorithms to output the optimal material ratio and molding process, and by completing model iteration updates and outputting full life cycle service prediction results, the system upgrades from static single-point evaluation to full-cycle dynamic accurate prediction, significantly improving the reliability and accuracy of medical biomaterial performance evaluation.

[0016] 2. Deeply explore the intrinsic evolution mechanism of material structure, biochemical environment, and service performance. This invention achieves in-depth understanding and mining of the multi-factor coupled degradation laws of medical biomaterials through a full-process digital twin modeling and time-series analysis mechanism, breaking through the limitations of traditional technologies that can only observe macroscopic performance changes and cannot analyze the internal coupling mechanisms. Through feature mapping fusion processing, it quantitatively reveals the coupling correlation mechanism between implant physical structural parameters and in vivo biochemical environment parameters, clarifying the intrinsic logic of multi-dimensional feature synergy. Through time-series simulation and sequence feature mining, it comprehensively sorts out the dynamic evolution time-series laws of material quality degradation and stress fatigue within a preset service cycle, deeply analyzing the progressive degradation mechanism of biomaterials in complex in vivo environments. Through iterative optimization of process parameters, it explores the matching relationship between material ratio, molding process, and dynamic service performance, clarifying the influence of process parameters on the material's anti-degradation and anti-fatigue performance. This invention achieves a comprehensive cognitive upgrade from macroscopic performance result judgment to microscopic time-series mechanisms and multi-factor coupling laws, improving the theoretical system of the full life-cycle performance evolution of medical biomaterials in complex in vivo environments, and providing in-depth mechanistic support for biomaterial structural design, ratio optimization, and molding process upgrades.

[0017] 3. Significantly improves the engineering practicality and scenario adaptability of the technical solution. This invention constructs a fully closed-loop intelligent optimization system encompassing "feature acquisition, feature fusion, twin modeling, dynamic simulation, pattern analysis, parameter optimization, and model iteration." It boasts strong engineering feasibility and wide applicability, adaptable to various medical implantable biomaterial R&D, optimization, and performance prediction scenarios. This invention is adaptable to implant products with different pore structures and material ratios, and can match the human implantation site environment with varying pH levels and biochemical characteristics, overcoming the drawbacks of traditional simulation models' fixed scenarios and poor versatility. By replacing traditional multiple in vitro and in vivo experiments with digital twin virtual simulation, it significantly reduces the cost of biomaterial R&D testing, shortens the R&D cycle, and avoids the problems of long physical testing cycles, poor repeatability, and high testing risks. Furthermore, this invention relies on intelligent iterative optimization algorithms to automatically generate optimal material ratios and molding process parameters, eliminating the need for trial-and-error based on human experience, lowering the threshold for process optimization, and improving the standardization and intelligence level of R&D. The updated target digital twin model can stably output prediction results of the entire life cycle service status, and can be directly applied to scenarios such as new material research and development, performance iteration of existing products, and clinical safety prediction. It is suitable for the industrial needs of large-scale research and development and refined performance optimization of medical biomaterials, and has extremely high industrialization and promotion value.

[0018] 4. Achieve closed-loop iterative optimization of material properties to ensure the long-term safety of medical implants. This invention establishes a continuous optimization closed-loop mechanism encompassing performance testing, deviation assessment, parameter optimization, and model iteration, overcoming the technical shortcomings of traditional biomaterial R&D, which involves "one-time modeling, one-time finalization, and no subsequent iterations." For material solutions that fail to reach the safe service threshold, this invention can automatically perform iterative optimization within the candidate process parameter space, accurately matching the optimal material ratio and molding process, and simultaneously updating the digital twin model to achieve dynamic matching and optimization of material structural parameters, process parameters, and service performance. Through dynamic prediction results throughout the entire lifecycle, potential risks such as material degradation failure and stress fatigue damage can be predicted in advance, enabling full-cycle controllability, predictability, and optimization of the service status of medical implants. This effectively improves the service stability and clinical safety of medical biomaterials, providing reliable technical support for the intelligent R&D and quality upgrade of high-end medical biomaterials. Attached Figure Description

[0019] Figure 1 This is a flowchart illustrating an embodiment of the method for simulating and optimizing the performance of medical biomaterials based on digital twins according to the present invention. Figure 2 This is a functional block diagram of an embodiment of the medical biomaterial performance simulation and optimization system based on digital twins of the present invention.

[0020] Explanation of icon numbers: 10. Biochemical environment feature set acquisition module; 20. Multi-dimensional fusion feature set acquisition module; 30. Initial digital twin model construction module; 40. Mass degradation sequence and stress fatigue sequence acquisition module; 50. Dynamic performance change feature determination module; 60. Target material ratio and target molding process acquisition module; 70. Target digital twin model acquisition module. Detailed Implementation

[0021] To better understand the above technical solutions, the following will provide a detailed explanation of the technical solutions in conjunction with the accompanying drawings and specific implementation methods.

[0022] like Figure 1 As shown, the first embodiment of this invention proposes a method for simulating and optimizing the performance of medical biomaterials based on digital twins, belonging to the technical fields of medical biomaterial simulation modeling, prediction of material performance time-series evolution, and intelligent optimization of process parameters. This invention addresses the technical pain points of traditional medical implant material development, which relies on physical sample fabrication, has a long cycle, high cost, cannot accurately reproduce the complex biochemical environment in vivo, is difficult to predict material degradation and stress-fatigue coupled failure, and relies on experience-based adjustment of process parameters. It proposes a fully intelligent solution encompassing morphological-biochemical dual-feature fusion, digital twin modeling, dynamic simulation throughout the entire service life, mining of time-series degradation features, performance threshold determination, iterative optimization of process parameters, and closed-loop model updates. This invention achieves accurate prediction of the full life-cycle performance of medical biomaterials and autonomous optimization of the molding process, significantly reducing fabrication costs, shortening the development cycle, and improving the safety and stability of implants during service. The steps include: Step S100: Obtain the set of physical morphological features corresponding to the target implant and the set of biochemical environmental features corresponding to the target implantation site; the set of physical morphological features includes pore distribution features, and the set of biochemical environmental features includes pH features.

[0023] This step collects the global physical morphological parameters of the target medical implant to be optimized, and constructs a complete set of physical morphological features; simultaneously, it collects the in vivo biochemical environment parameters of the target implantation site in the human body, and constructs a set of biochemical environment features; thus realizing the dual-dimensional raw data input of "implant body structural features + human service environment features", providing a real, standardized and quantifiable basic dataset for subsequent feature fusion modeling and performance simulation.

[0024] Target implants refer to medical biomaterial products that can be implanted into the human body for tissue repair, bone support, soft tissue filling, and vascular repair. They include biodegradable polymer materials, composite bioceramics, and biodegradable metal materials. The size range of implants is 0.5mm to 500mm, which is compatible with the size of conventional clinical implantable devices.

[0025] The physical morphological feature set is a multidimensional feature dataset that quantitatively characterizes the macroscopic structure and microscopic morphology of implants. Its core includes pore distribution features, while also encompassing pore size, porosity, pore connectivity, wall thickness, specific surface area, structural curvature, and dimensional characteristics. The numerical range and criteria for pore distribution features are as follows: 1. Average pore size: 20μm~800μm, preferably 50μm~400μm; 2. Porosity: 30%~90%, preferably 45%~75%; 3. Porosity: ≥85%; Criteria for these values: This range represents the optimal pore parameter range for biodegradable medical implant materials. Pores that are too small are detrimental to cell adhesion and fluid permeation, while pores that are too large result in insufficient structural mechanical strength, failing to meet the requirements for long-term service support.

[0026] The target implantation site is the tissue or bone area in the human body to be implanted, including bone tissue, subcutaneous soft tissue, vascular tissue, and joint tissue. Different sites correspond to different biochemical environments with different pH levels, ion concentrations, temperatures, and enzyme activities.

[0027] The biochemical environmental characteristic set is a multidimensional set of parameters that quantitatively characterize the microenvironment within the implantation site in the human body. Its core includes pH value, while also encompassing body temperature, ion concentration, body fluid osmotic pressure, degrading enzyme activity, and tissue fluid flow rate. The pH value range and its basis are as follows: Normal human tissue pH steady-state range: 6.8–7.4; inflammatory state fluctuation range: 6.5–7.8; The simulation range used in this method is 6.5–7.8; The basis for these values ​​is to fully cover both normal physiological states and mild postoperative inflammatory pathological states, realistically recreating the complex acid-base corrosion environment within the implanted object during service.

[0028] Step S200: Use a classification regression algorithm to perform feature mapping and fusion processing on the physical morphology feature set and the biochemical environment feature set to obtain a multidimensional fusion feature set; each multidimensional fusion feature set corresponds to a unique combination relationship between the implant and the implantation environment.

[0029] Using the physical morphology feature set and biochemical environment feature set obtained in step S100 as dual input sources, a classification and regression algorithm is used to complete the nonlinear mapping, correlation coupling and weighted fusion of cross-dimensional features, eliminate feature dimension differences, redundant features and invalid noise features, and explore the coupling relationship between structural parameters and environmental parameters to generate a high-dimensional, orthogonal and standardized multidimensional fusion feature set. Each set of multidimensional fusion feature sets uniquely corresponds to a combination of "implant structure morphology + in vivo biochemical environment" working conditions, providing standardized modeling input for the digital twin model.

[0030] The classification and regression algorithm (the core algorithm architecture of this step) adopts a multi-feature nonlinear classification and regression fusion algorithm with a five-layer architecture, specifically designed for the fusion of morphological and biochemical heterogeneous features: 1. Feature normalization layer: unifies the mapping of physical morphology and biochemical environment features of different dimensions to the [0,1] interval; 2. Correlation screening layer: calculates the Pearson correlation coefficient of features and removes redundant and invalid features with a correlation coefficient ≤0.1; 3. Nonlinear mapping layer: completes the mapping of linear features to a high-dimensional nonlinear space through a polynomial kernel function to fit the material-environment coupling relationship; 4. Weighted fusion layer: adaptively allocates weights according to feature sensitivity, with higher weights for highly sensitive features; 5. Feature output layer: generates a multi-dimensional fusion feature set with fixed dimensions. Algorithm parameter thresholds: feature retention threshold correlation coefficient >0.1, fusion fitting accuracy R²≥0.992; value basis: a low correlation coefficient indicates that the feature has no significant impact on material performance, and removing it can reduce the computational load of the model and improve the modeling accuracy.

[0031] Feature mapping fusion processing refers to the quantitative processing process that couples the physical features of a structure with the features of its biochemical environment in vivo across domains to establish a correlation mapping relationship between "structural parameters, environmental parameters, and material properties," thereby solving the problems of decoupling between the structure and the environment and simulation distortion in traditional modeling.

[0032] The multidimensional fusion feature set is a standardized high-dimensional feature vector set that integrates structural morphology and biochemical environment. The dimension of a single feature set is 32 to 128, with 64 dimensions being preferred. The selection criteria are: too low a dimension cannot represent complex coupling relationships, while too high a dimension causes model overfitting and computational redundancy. 64 dimensions take into account both completeness and stability.

[0033] The combination relationship between implant and implantation environment is a unique combination of working conditions corresponding to each set of fusion features, realizing a precise twin correspondence of one object and one model, and one environment and one model, avoiding the defects of insufficient simulation accuracy of general models.

[0034] Step S300: Construct an initial digital twin model based on a multi-dimensional fusion feature set. The initial digital twin model is used to simulate the material evolution process of the target implant in a real physical environment.

[0035] Using the multidimensional fusion feature set generated in step S200 as the modeling driving parameters, and combining the constitutive parameters, thermodynamic parameters, degradation kinetic parameters, and mechanical fatigue parameters of medical biomaterials, an initial digital twin model of structure-environment-performance multi-field coupling is constructed. This model realizes the digital replication of the real human physical, chemical, and mechanical environment, and can simulate the dynamic evolution behavior of the target implant in the real physical environment, including material corrosion, degradation, mechanical fatigue, and structural evolution, with high precision.

[0036] The initial digital twin model is a digital twin simulation model that maps one-to-one with the real target implant and the real implantation environment. It includes four sub-modules: geometric twin, environmental twin, material property twin, and mechanical response twin. The model's simulation accuracy is ≥99%, and the error with the actual test sample is ≤3%. Model mesh parameter thresholds: The number of model mesh elements is 100,000 to 2 million, adaptively divided according to the implant size; the mesh quality pass rate is ≥98%. The criteria for the values ​​are: too few mesh elements result in simulation distortion, and too many elements result in computational redundancy. This range is suitable for the simulation needs of fine medical structures.

[0037] The material evolution process is a comprehensive dynamic process of mass degradation, structural loosening, mechanical attenuation, stress concentration, fatigue damage, and micropore evolution that occurs in the implant under the biochemical environment in the body. It is the core simulation object of this initial digital twin model.

[0038] Step S400: Dynamically simulate the service process of the target implant within the preset service period using the initial digital twin model to obtain the mass degradation sequence and stress fatigue sequence corresponding to the target implant.

[0039] Based on the completed initial digital twin model, a preset service cycle conforming to clinical medical standards was set, and multi-physics field coupled dynamic time-series simulation was carried out. The mass loss change data and mechanical stress fatigue change data of the implant were solved step by step over time, and stored sequentially according to the time axis to generate the mass degradation time-series sequence and stress fatigue time-series sequence corresponding to the target implant, respectively, providing time-series data support for subsequent performance evolution law mining.

[0040] The preset service period is the standard service simulation duration of medical implants, ranging from 3 months to 36 months, with 6 months to 24 months being preferred. The value is based on the clinical service period of short-term biodegradable implants and medium- to long-term support implants, and complies with the evaluation standards for biomaterials of medical devices.

[0041] Dynamic simulation time step threshold: simulation step size 0.5 days to 3 days, preferably 1 day; the basis for the value is: if the step size is too large, the subtle evolution patterns will be lost, if the step size is too small, the amount of computation will increase dramatically, and a step size of 1 day takes into account both timing accuracy and simulation efficiency.

[0042] The mass degradation sequence is a time-series array of the percentage of remaining implant mass over time, with values ​​ranging from [0, 100%], characterizing the changes in material degradation rate and residual mass.

[0043] The stress fatigue sequence is a time-based array of the equivalent fatigue stress and cumulative cyclic damage of the implant, representing the mechanical fatigue attenuation caused by micromotion, stress, and tissue tension in vivo. The stress calculation accuracy is ≤0.01MPa.

[0044] Step S500: Input the mass degradation sequence and stress fatigue sequence into the sequence analysis model, extract the multidimensional time-series evolution law, and determine the dynamic performance change characteristics of the target implant; the dynamic performance change characteristics are used to characterize the degradation trend of the target implant material in complex environments.

[0045] The mass degradation sequence and stress fatigue sequence obtained in step S400 are input into the trained sequence analysis model. Multi-dimensional time-series features such as degradation rate time-series law, mechanical decay time-series law, abrupt damage law, and steady-state evolution law are automatically extracted and quantified to obtain the dynamic performance change features of the target implant. The dynamic performance change features are used to accurately characterize the long-term degradation trend and failure risk of medical biomaterials in the complex acid-base and mechanical coupling environment in vivo.

[0046] Sequence Analysis Model (Complete Architecture of Core Algorithm in this Step): Employs a temporal attention mechanism-LSTM deep sequence analysis model with a six-layer architecture: 1. Temporal Data Input Layer: Receives dual-path temporal sequences of quality degradation and stress fatigue; 2. Temporal Smoothing Preprocessing Layer: Removes minor noise fluctuations from the simulation; 3. LSTM Temporal Encoding Layer: Captures long-period dependencies and slow degradation patterns; 4. Attention Weight Layer: Enhances degradation and abrupt change features at key time nodes; 5. Multidimensional Feature Extraction Layer: Outputs four types of features: degradation, mechanical, stability, and risk; 6. Performance Feature Fusion Layer: Generates the final dynamic performance change features. Model Performance Parameters: Temporal fitting accuracy ≥ 0.99, feature extraction error ≤ 0.5%.

[0047] The multidimensional time-series evolution law includes five core evolution modes: uniform degradation law, accelerated degradation law, stress accumulation law, fatigue saturation law, and performance steady state law, covering the failure modes of the entire material life cycle.

[0048] Dynamic performance change characteristics are multi-dimensional feature vectors that quantitatively characterize the comprehensive service performance of materials. They include five quantitative indicators: degradation uniformity, mechanical retention rate, fatigue damage degree, performance stability, and failure risk coefficient. The value range is [0,1]. The larger the value, the better the performance and the lower the risk.

[0049] The degradation trend is the overall change trend of implant materials in terms of mass loss, mechanical decline, structural failure, and performance degradation as the service time increases. It is the core basis for determining whether a material meets clinical safety standards.

[0050] Step S600: If the dynamic performance change characteristics do not reach the preset safe service performance threshold, then the dynamic performance change characteristics are used as input parameters, and the parameter optimization algorithm is used to perform iterative optimization processing in the preset candidate process parameter space to obtain the target material ratio and the target molding process.

[0051] A preset clinical safety service performance threshold for medical implants is established. The dynamic performance change characteristics obtained in step S500 are compared with the threshold for judgment. If the dynamic performance change characteristics do not reach the preset safety service performance threshold, it is determined that the current material ratio and molding process cannot meet the long-term clinical service safety requirements. The parameter optimization algorithm is started to perform global iterative optimization of the material ratio parameters and molding process parameters in the preset legal candidate process parameter space, and the optimal combination of target material ratio and target molding process parameters is obtained by convergence solution.

[0052] Preset safe service performance thresholds: 1. Mechanical retention rate threshold ≥ 85%; 2. Degradation uniformity threshold ≥ 0.8; 3. Fatigue damage threshold ≤ 0.15; 4. Failure risk coefficient threshold ≤ 0.1; Unified value basis: Complies with "GB / T 16886 Biological Evaluation of Medical Devices" and clinical safety standards for biodegradable medical implant materials. If the value is lower than this threshold, the material is considered to have the risk of fracture, premature degradation, and mechanical failure, and does not meet the implant service requirements. The process must be optimized.

[0053] The candidate process parameter space is a preset range of adjustable material parameters, including: 1. Material ratio range: polymer substrate ratio 50%~95%, functional modification component 5%~50%; 2. Molding process range: sintering temperature 800℃~1300℃, molding pressure 2MPa~20MPa, holding time 0.5h~4h; The selection criteria are: covering the conventional molding process window of medical biomaterials to ensure that the optimization results are engineering feasible.

[0054] Parameter optimization algorithm (complete architecture of the core algorithm in this step): An adaptive genetic iterative optimization algorithm with a five-layer architecture is adopted: 1. Population initialization layer: Randomly generate an initial process parameter population within the parameter space; 2. Fitness calculation layer: Use dynamic performance characteristics as the fitness evaluation index; 3. Selection, crossover, and mutation layer: Adaptively update the iterative parameters; 4. Constraint screening layer: Remove parameters that exceed the process window or have no engineering significance; 5. Convergence determination layer: Convergence is determined when the fitness change is ≤0.001 for 10 consecutive generations, and the optimal process is output. Iteration threshold basis: Convergence error ≤0.001, ensuring the accuracy of the optimal process parameters and avoiding local optima.

[0055] The target material ratio and target molding process are the globally optimal process combination obtained after iterative convergence, which enables the material degradation compatibility, mechanical retention and fatigue resistance to reach the optimal state for clinical safety.

[0056] Step S700: Update the initial digital twin model based on the target material ratio and target molding process to obtain the target digital twin model, and output the full life cycle service status prediction results of the target implant through the target digital twin model.

[0057] The target material ratio parameters and target molding process parameters obtained from step S600 are used to back-update the material constitutive parameters, structural parameters, mechanical parameters and degradation kinetic parameters of the initial digital twin model, and reconstruct the multi-field coupling relationship within the model to obtain a high-precision target digital twin model corresponding to the optimal process. Based on the target digital twin model, the full life cycle simulation is completed, and the prediction results of the implant's full life cycle service status from the initial implantation stage, steady-state service period to final complete degradation are output, realizing accurate optimization and long-term prediction of material performance.

[0058] The target digital twin model is a high-precision twin model that has been updated through closed-loop optimization of process parameters. The model performance prediction error is ≤1.5%, and the accuracy is improved by ≥40% compared with the initial model. It can accurately characterize the actual service behavior of materials under the optimal process.

[0059] The full life cycle service status prediction results are a complete set of quantitative results covering the initial implantation adaptability, mid-term mechanical stability, late-term degradation consistency, and final complete degradation cycle. These results include degradation cycle duration, mechanical attenuation curve, fatigue life, structural integrity, and biocompatibility indicators, and can be directly used for medical material process finalization and clinical safety assessment.

[0060] The present invention provides a method for simulating and optimizing the performance of medical biomaterials based on digital twins. It constructs an intelligent optimization system for medical biomaterials, which includes "fusion of structural and biochemical dual features, digital twin modeling, full-cycle dynamic simulation, temporal degradation mining, performance threshold determination, process iteration optimization, and model closed-loop update". This method breaks through the traditional physical trial-and-error R&D model, realizes virtual evaluation of material performance, autonomous process optimization, and accurate prediction of the entire life cycle. It significantly improves the service safety, degradation compatibility, and mechanical stability of medical implant materials, greatly shortens the R&D cycle, reduces R&D costs, and has extremely high clinical application value and engineering promotion value.

[0061] Furthermore, the method for simulating and optimizing the performance of medical biomaterials based on digital twins provided in this embodiment includes step S100 as follows: Step S110: Extract the set of physical morphological features corresponding to the target implant.

[0062] The formula for calculating the physical morphological feature vector is: (1) In formula (1), This is a physical morphological feature vector, without units, derived from the CT / MRI three-dimensional scan data of the implant, and takes the value of a three-dimensional real number vector, used to uniformly characterize the overall physical morphological parameters of the implant; Porosity is a unitless property derived from porosity statistics in scanned data. Its value ranges from 0.1 to 0.9, representing the proportion of pores in the total volume of the implant. The pore size distribution density is expressed in units of pores / mm³. It is derived from the pore size statistical calculation of the scanned data and ranges from 10 to 500 pores / mm³, representing the number of pores per unit volume. Surface roughness, in μm, is derived from surface morphology detection equipment and ranges from 0.5 to 20 μm, characterizing the microscopic unevenness of the implant surface. The control logic of formula (1) is to perform three-dimensional scanning and surface morphology detection on the target implant, extract three core physical parameters: porosity, pore size distribution density, and surface roughness, and integrate them into a standardized three-dimensional physical morphology feature vector in a preset order, thereby realizing the quantitative and structural characterization of the physical morphology of the implant and providing complete basic data support for subsequent mechanical coupling and biological response simulation. Traditional technology only collects the porosity parameter, which cannot reflect the key influence of pore size distribution density and surface roughness on material degradation behavior and tissue cell adhesion and bonding effect. The physical morphology characterization is singular and information is seriously lacking. Formula (1) constructs a three-dimensional physical morphology feature vector containing porosity, pore size distribution density, and surface roughness, which comprehensively covers the internal pore structure and surface micromorphology features of the implant, providing basic data with complete data dimensions and higher characterization accuracy for subsequent mechanical environment coupling, biological response simulation and optimization design, and significantly improving the accuracy and reliability of subsequent performance simulation.

[0063] The medical biomaterial performance simulation and optimization system first extracts a set of physical morphological features corresponding to the target implant. This set of features characterizes the implant's microstructural parameters, specifically including core structural indicators such as implant porosity and surface roughness. The system further acquires a set of host biochemical environment features corresponding to the target implantation site. This set includes in vivo microenvironment parameters such as local pH and body fluid flow shear stress.

[0064] Step S120: Based on the set of physical morphological features, obtain the set of biochemical environmental features corresponding to the target implantation site.

[0065] The formula for calculating the biochemical environmental feature vector is: (2) In formula (2), This is a biochemical environment feature vector, without units, derived from tissue fluid sampling and fluid dynamics analysis at the implantation site. Its value is a three-dimensional real number vector used to uniformly characterize the biochemical and mechanical coupling environment of the implantation site. The pH value of the tissue fluid is unitless, derived from a pH measuring instrument, and ranges from 5.5 to 8.0, characterizing the acid-base environment of the implantation site. The active ion concentration, in mmol / L, is derived from ion concentration detection equipment and ranges from 0.1 to 5.0 mmol / L. It characterizes the content of active ions in tissue fluid that accelerate material degradation. The local fluid shear stress, in Pa, is derived from fluid dynamics simulation and actual measurement, with a value range of 10~1000 Pa, representing the continuous scouring force of body fluid flow on the implant. The control logic of formula (2) is to collect the biochemical parameters and fluid dynamic parameters of the tissue fluid at the target implantation site, and integrate the three core indicators of pH, active ion concentration, and local fluid shear stress into a standardized three-dimensional environmental feature vector in a preset order, constructing a biochemical-mechanical coupled representation of the real service environment of the implantation site, providing accurate environmental input for the subsequent coupled simulation of material degradation and mechanical response. Traditional technology only detects a single biochemical indicator of pH, which cannot reflect the synergistic effect of body fluid shear stress on material wear and degradation behavior, and the environmental representation is singular and deviates greatly from the real human service environment. Formula (2) integrates two key environmental parameters, biochemical and mechanical, to construct a multi-dimensional coupled environmental feature vector, more realistically restoring the complex service environment of the human implantation site, and significantly improving the accuracy and clinical reference value of subsequent material performance simulation and optimization results.

[0066] The medical biomaterial performance simulation and optimization system performs cross-correlation analysis on physical morphological characteristics and biochemical environmental characteristics to establish feature mapping relationships: it couples the implant porosity with the shear force of body fluid flow, and simultaneously matches the implant surface roughness with local pH. Based on multidimensional feature mapping rules, a degradation kinetic analysis model is constructed. This degradation kinetic analysis model is used to simulate the permeation, diffusion, and erosion processes of body fluid in the implant's microporous structure, accurately predicting the initial degradation rate of the implant under the current host biochemical environment.

[0067] Step S130: If the initial degradation rate of the material corresponding to the biochemical environment feature set is greater than the preset degradation threshold, then collect the target ion concentration.

[0068] The formula for calculating and determining the initial degradation rate of the material is as follows: (3) (4) In formulas (3)~(4), The initial degradation rate of the material, in Pa / s, is derived from the result of a two-factor weighted calculation and is a positive real number, representing the rate of change in the strength of the material under biochemical erosion per unit time. The coefficient representing the influence of ion degradation is dimensionless and is derived from in vitro degradation calibration experiments of the material. Its value ranges from 0.2 to 1.5, characterizing the influence weight of active ions in tissue fluid on the degradation rate of the material. The active ion concentration, in mmol / L, is derived from the biochemical environment feature vector in step S120 and characterizes the content of active ions in the tissue fluid that accelerate material degradation. The acid-base degradation influence coefficient is a unitless coefficient derived from in vitro degradation calibration experiments of materials, with a value range of 0.1 to 1.0. It characterizes the weight of the influence of acid-base deviation from human physiological standard values ​​on the degradation rate. The pH value of the tissue fluid is unitless and is derived from the biochemical environment feature vector in step S120, representing the acid-base environment of the implantation site. The preset degradation rate threshold, in Pa / s, is derived from safety experimental standards and ranges from 0.1 to 1.0 Pa / s. It serves as the criterion for triggering refined ion collection. The ion acquisition trigger marker has no unit and takes the value of 0 or 1, where 1 indicates that the target ion concentration fine acquisition process is executed and 0 indicates that no additional acquisition is required. The control logic of formulas (3) to (4) is to combine the active ion concentration and pH parameters output in step S120, and perform a two-factor weighted calculation by the ion degradation influence coefficient and the pH degradation influence coefficient to calculate the initial degradation rate of the material at the target implantation site; the calculated initial degradation rate is compared with the preset safety threshold. When the degradation rate exceeds the threshold, the fine target ion concentration acquisition process is triggered, providing more accurate ion environment data support for subsequent degradation mechanism analysis and material optimization. Traditional technology uses a fixed single threshold to determine the material degradation risk, without distinguishing the different effects of active ions and pH on the degradation rate. The judgment logic is crude and cannot reflect the real in vivo degradation mechanism. Formulas (3) to (4) calculate the initial degradation rate by ion and pH two-factor weighted calculation, construct a judgment model that fits the in vivo degradation mechanism of biomaterials, greatly improve the scientificity and accuracy of degradation risk judgment, and provide a reliable basis for fine environmental analysis and optimization of materials.

[0069] If the initial degradation rate calculated by the degradation kinetics analysis model is greater than the preset degradation rate threshold, the medical biomaterial performance simulation and optimization system collects real-time data on the concentration of target metal ions released during the implant degradation process. The system extracts the temporal gradient of the target ion concentration and, combined with the cell tolerance concentration range corresponding to the local human tissue, quantifies and solves the target ion concentration matching weight.

[0070] Step S140: Calculate the fusion matching degree value based on the target ion concentration.

[0071] The formula for calculating the material-environment fusion compatibility is: (5) In formula (5), The material-environment fusion matching degree is dimensionless and derived from the result of normalized weighted calculation. The value ranges from 0 to 1, representing the degree of compatibility between the implant material and the biochemical environment of the implantation site. The larger the value, the worse the compatibility between the material and the environment. This is the ion characteristic weighting coefficient, dimensionless, derived from in vitro degradation experimental calibration, with a value range of 0.3~0.7, and satisfying the following conditions: This is used to adjust the proportion of the influence of active ion concentration on the matching degree calculation; The target ion concentration, in mmol / L, is derived from the refined acquisition results in step S130 and characterizes the actual content of active ions in the tissue fluid at the implantation site. The standard reference ion concentration, in mmol / L, is derived from the standard parameters of normal human tissue fluid and is fixed at 1.0 mmol / L, providing a benchmark for ion concentration normalization. The acid-base characteristic weighting coefficient is dimensionless, derived from in vitro degradation experiments, and ranges from 0.3 to 0.7. It is used to adjust the proportion of the influence of the acid-base parameter on the matching degree calculation. The pH value of the tissue fluid is unitless and is derived from the biochemical environment feature vector in step S120, representing the actual acid-base environment of the implantation site. The standard pH deviation reference value is unitless, derived from the standard human physiological parameters, and is fixed at 0.6, representing the normal pH fluctuation range of human tissue fluid, providing a benchmark for pH deviation normalization. The control logic of formula (5) is based on the standard human physiological parameters, normalizing the active ion concentration and pH deviation of the actual implantation site, and then performing weighted calculations by combining the preset ion characteristic weight coefficient and the pH characteristic weight coefficient to quantitatively calculate the fusion matching degree between the implant material and the biochemical environment of the implantation site, realizing the standardization and quantitative characterization of the material-environment adaptation degree, and providing accurate quantitative basis for subsequent material performance simulation and optimization. Traditional technology relies solely on experience to qualitatively judge the compatibility between biomaterials and the implantation environment, lacking a unified and objective quantitative judgment standard. The compatibility assessment is highly subjective and has low precision, failing to provide reliable data support for material optimization design. Formula (5) constructs a quantitative calculation model for the material-environment fusion matching, and achieves accurate quantitative characterization of the degree of fit through normalized weighted operation. This provides a directly usable quantitative basis for subsequent material performance simulation, degradation behavior prediction and structural optimization, and significantly improves the scientificity and repeatability of material optimization design.

[0072] The deviation between the target ion concentration matching weight and the initial degradation rate is weighted and fused to obtain the fusion matching degree between the implant and the host environment.

[0073] Step S150: Based on the fusion matching degree value, obtain the set of physical morphological features corresponding to the target implant and the set of biochemical environmental features corresponding to the target implantation site.

[0074] The formula for calculating the integrated feature vector is: (6) In formula (6), To integrate feature vectors, which are unitless and derived from vector concatenation operations, with values ​​being real number vectors, this dataset is used to integrate three core parameters: physical morphology, biochemical environment, and material-environment fusion matching degree, serving as the basic dataset for subsequent material performance simulation and optimization. This is a physical morphological feature vector, without units, derived from the output of step S110, representing the physical morphological parameters of the implant, such as porosity, pore size distribution density, and surface roughness. This is a biochemical environment feature vector, without units, derived from the output of step S120, characterizing biochemical and mechanical environmental parameters such as pH, active ion concentration, and local fluid shear stress at the implantation site. The material-environment fusion matching degree is a unitless value derived from the output of step S140. It characterizes the degree of compatibility between the implant material and the biochemical environment of the implantation site. The larger the value, the worse the compatibility. The control logic of formula (6) is to structurally splice the physical morphology feature vector output from step S110, the biochemical environment feature vector output from step S120, and the material-environment fusion matching degree calculated from step S140 in a preset dimensional order to form a complete integrated feature vector containing three types of information: material morphology, service environment, and compatibility assessment. This provides multi-dimensional and integrated basic data support for subsequent material performance simulation, degradation behavior prediction, and optimization design based on digital twins. The feature set constructed by traditional technology only contains the basic morphology of the material and environmental parameters. It does not introduce the key coupling index of material-environment compatibility and cannot reflect the interaction between the material and the implantation environment. Subsequent simulation analysis may easily ignore the influence of environmental compatibility on material degradation and mechanical response. Formula (6) incorporates the fusion matching degree as an independent feature into the integrated dataset, realizing a unified representation of material morphology, environmental parameters and adaptability assessment, strengthening the ability of subsequent feature fusion and performance simulation to express the material-environment coupling relationship, and significantly improving the accuracy of simulation results and the pertinence of optimization design.

[0075] The medical biomaterial performance simulation and optimization system uses a fusion matching degree numerical method to reverse-correct the correlation mapping between physical morphological characteristics and biochemical environmental characteristics, enabling precise control of the implant degradation rate. This embodiment can effectively match the host tissue tolerance characteristics, inhibiting problems such as excessively rapid degradation and excessive ion concentration, significantly improving the biocompatibility and in vivo stability of the implant during long-term implantation.

[0076] Preferably, the method for simulating and optimizing the performance of medical biomaterials based on digital twins provided in this embodiment includes step S200: Step S210: Extract the surface roughness feature values ​​contained in the physical morphology feature set and the local shear stress feature values ​​contained in the biochemical environment feature set.

[0077] The formulas for calculating the characteristic values ​​of surface roughness and local shear stress are as follows: (7) In formula (7), The surface roughness feature value, in μm, is directly taken from formula (1). This is used for subsequent rejection risk classification calculations; To extract the local shear stress characteristic value, in Pa, it is directly taken from formula (2). This is used for subsequent rejection risk classification calculations; The surface roughness of the target implant, in μm, is derived from the physical morphology feature vector in step S110 and characterizes the degree of microscopic unevenness on the implant surface. The local fluid shear stress at the target implantation site, in Pa, is derived from the biochemical environment feature vector in step S120 and represents the continuous scouring force of body fluid flow on the implant. The control logic of formula (7) is to accurately screen and extract two core parameters strongly related to tissue rejection reaction and interface wear behavior from the physical morphology feature vector constructed in step S110 and the biochemical environment feature vector constructed in step S120—surface roughness and local shear stress—to form a set of low-redundancy, high-correlation feature variables as input for the subsequent implant rejection risk classification model. Traditional rejection risk modeling often directly uses all original features, resulting in high data redundancy and large interference from irrelevant features, leading to low model computation efficiency and poor generalization ability. Based on the mechanism analysis of biomaterial implantation, formula (7) accurately screens key features strongly related to rejection and wear, eliminates redundant and irrelevant data, and significantly reduces the input dimension of the subsequent classification model, which not only improves computation efficiency but also strengthens the model's learning ability of core influencing factors, making risk classification more targeted and accurate.

[0078] The medical biomaterial performance simulation and optimization system extracts surface roughness parameters from the physical morphological feature set and local fluid shear stress parameters from the biochemical environment feature set. Based on these two core parameters, it performs sample node partitioning and risk classification. The node partitioning process is a standardized classification process that maps continuous physical parameters to a discrete evaluation space.

[0079] Step S220: Based on the surface roughness characteristic value and local shear stress, perform node division processing to obtain the initial classification label of rejection risk.

[0080] The formula for calculating the initial classification label of rejection risk is: (8) In formula (8), The initial classification label for rejection risk is unitless, derived from the output of the symbolic function, and takes the value of +1 or -1, where +1 represents a high rejection risk label and -1 represents a normal adaptation label. This is a sign function with no unit, derived from standard mathematical operations. It outputs +1 when the input value is greater than 0 and -1 when the input value is less than or equal to 0. It is used to output standardized binary classification labels. The roughness classification coefficient is dimensionless and comes from the training and calibration of the rejection risk classification model. Its value ranges from 0.01 to 0.5, representing the weight of the influence of surface roughness on the tissue rejection risk. The surface roughness feature value, in μm, is derived from the extraction result in step S210 and characterizes the degree of microscopic unevenness on the implant surface. The shear stress classification coefficient is dimensionless and is derived from the training and calibration of the rejection risk classification model. Its value ranges from 0.001 to 0.1, representing the weight of the influence of local fluid shear stress on the tissue rejection risk. The local shear stress characteristic value, in Pa, is derived from the extraction result of step S210 and characterizes the continuous scouring force of body fluid flow on the implant. The threshold for rejection risk assessment is unitless, derived from clinical trial data, and ranges from 5 to 50, serving as the critical value for rejection risk classification. The control logic of formula (8) is to linearly weight the surface roughness feature value and local shear stress feature value extracted in step S210 using the roughness classification coefficient and shear stress classification coefficient, respectively, and then calculate the difference with the preset rejection risk assessment threshold. Finally, a standardized binary classification label for rejection risk is output through a sign function, realizing rapid assessment and classification of implant tissue rejection risk. Traditional rejection risk assessment relies solely on the single morphological parameter of material surface roughness, ignoring the synergistic effect of body fluid shear stress, a key fluid dynamics factor, on interface wear and inflammatory response. The assessment logic is one-sided and the accuracy is low. Formula (8) constructs a two-parameter linear classification model and introduces surface roughness and local shear stress for joint risk assessment, fully considering the synergistic effect of material morphology and fluid dynamics environment, significantly improving the accuracy and clinical reference value of rejection risk classification.

[0081] The medical biomaterial performance simulation and optimization system constructs a two-dimensional feature evaluation coordinate system, with implant surface roughness as the horizontal axis and local body fluid shear stress as the vertical axis. A decision tree algorithm is used to finely divide the two-dimensional feature space into grids, with each grid node corresponding to a unique tissue response mode and risk level. The system then substitutes the real-time extracted surface roughness and local shear stress feature values ​​into the two-dimensional feature evaluation coordinate system to locate target nodes and assigns corresponding initial rejection risk classification labels based on node attributes.

[0082] Step S230: If the initial classification label is a high-risk rejection label, collect the concentration of degradation products and the osmotic pressure of tissue fluid, and perform fitting calculation on the concentration of degradation products and the osmotic pressure of tissue fluid to obtain the feature mapping weight matrix.

[0083] The formula for calculating the feature mapping weight matrix is: (9) (10) In formulas (9)~(10), This is an environmental response observation vector, without units, derived from sample collection and structured integration. Its value is a two-dimensional real vector containing two key environmental response parameters: degradation product concentration and tissue fluid osmotic pressure. The concentration of degradation products is expressed in mg / L, derived from biochemical detection equipment, and ranges from 0 to 20 mg / L, characterizing the content of harmful substances released during the material degradation process. The value is the osmotic pressure of tissue fluid, expressed in kPa. It is derived from osmotic pressure testing equipment and ranges from 70 to 100 kPa. It characterizes the osmotic state of tissue fluid at the implantation site. The basic feature sample matrix is ​​unitless and consists of multiple sets of physical morphological feature vectors. Biochemical environmental feature vectors The sample composition serves as the matrix of independent variables for linear fitting; The feature mapping weight matrix is ​​a unitless matrix derived from the least squares fitting result. It is a real number matrix used to achieve adaptive mapping between the physical characteristics of the implant and the environmental response characteristics of the implantation site. The control logic of formulas (9) to (10) is that when step S220 determines that the initial classification label is a high rejection risk label, the refined environmental parameter acquisition process is triggered to obtain the concentration of degradation products and tissue fluid osmotic pressure of the target implantation site and construct a standardized environmental response observation vector. The basic feature sample matrix is ​​constructed using multiple sets of physical morphology and biochemical environmental feature samples collected in history. The least squares method is used to linearly fit the environmental response observation vector to obtain the feature mapping weight matrix, thereby achieving accurate correlation mapping between material characteristics and environmental response in high-risk scenarios. Traditional feature fusion methods use a fixed weight matrix, which cannot adapt to the dynamic changes in the correlation between material and environmental characteristics in high rejection risk scenarios, and the mapping model has poor generalization. Formulas (9) to (10) are based on measured environmental response data. By adaptively solving the feature mapping weight matrix using the least squares method, they can accurately capture the feature correlation changes in high-risk scenarios, realize the differentiated and scenario-based mapping of material-environment features, and significantly improve the accuracy and pertinence of subsequent performance simulation and risk assessment.

[0084] If the initial classification label for rejection risk is determined to be high-risk, the medical biomaterial performance simulation and optimization system simultaneously collects data on the concentration of implant degradation products and the osmotic pressure of tissue fluid at the implantation site. Based on a multinomial regression algorithm, the nonlinear correlation between the accumulation of degradation products and changes in tissue fluid osmotic pressure is explored. A multiple regression equation is constructed with the concentration of degradation products as the independent variable. Sensitivity coefficients for each dimension parameter are extracted by solving the partial derivatives of the equation, and a standardized mapping weight matrix is ​​constructed based on the arrangement of these sensitivity coefficients.

[0085] Step S240: Use a mapping weight matrix to perform feature mapping fusion processing on the physical morphology feature set and the biochemical environment feature set to obtain a multi-dimensional fused feature set.

[0086] The formula for calculating the multidimensional fusion feature set is: (11) In formula (11), It is a multidimensional fusion feature set, without units, derived from the result of matrix multiplication, and takes the value of a real number vector. It is a standardized feature after the fusion of physical morphological features and biochemical environment features. Each vector corresponds to a unique implant-implantation environment combination. This is the feature mapping weight matrix, which has no unit and is derived from the least squares fitting result in step S230. It is used to achieve adaptive mapping between physical features and environmental features. This is a physical morphological feature vector, without units, derived from the output of step S110, representing the physical morphological parameters of the implant, such as porosity, pore size distribution density, and surface roughness. The biochemical environment feature vector is dimensionless and originates from the output of step S120. It represents the biochemical and mechanical environmental parameters of the implantation site, such as pH, active ion concentration, and local fluid shear stress. The control logic of formula (11) is to use the feature mapping weight matrix adaptively solved in step S230 to linearly map and fuse the physical morphology feature vector constructed in step S110 and the biochemical environment feature vector constructed in step S120. This integrates the material physical morphology information and the implantation environment information into a unified multi-dimensional fusion feature vector, providing standardized and integrated feature input for subsequent material performance simulation and risk assessment based on digital twins. Traditional feature fusion uses fixed mapping weights, which cannot reflect the differences in the correlation between material and environmental features under different risk levels. The fusion features have poor adaptability to high-risk scenarios. Formula (11) adopts a risk-adaptive fusion strategy, dynamically adjusting the feature mapping relationship based on measured data under high rejection risk scenarios. The generated fusion features can more accurately capture the potential trends of material degradation and tissue rejection, significantly improving the accuracy and pertinence of subsequent performance simulation and risk assessment.

[0087] The medical biomaterial performance simulation and optimization system uses a mapping weight matrix to perform matrix multiplication operations on the original physical morphological features and biochemical environmental features, completes the projection mapping and dimension unification of multi-dimensional features, eliminates the dimensional differences of different parameters, and finally generates a multi-dimensional fusion feature set containing structural characteristics, environmental characteristics, and degradation characteristics, providing standardized input for subsequent mechanical simulation and state modeling.

[0088] Furthermore, the method for simulating and optimizing the performance of medical biomaterials based on digital twins provided in this embodiment includes step S300 as follows: Step S310: Analyze the multidimensional fusion feature set and extract spatial geometric features and material property features.

[0089] The formula for splitting the dimensions of a multidimensional fusion feature set is: (12) In formula (12), It is a multi-dimensional fusion feature set, without units, derived from the feature mapping fusion result of step S240, and takes the value of a real number vector. It is a standardized feature after the fusion of physical morphological features and biochemical environmental features. It is a spatial geometric feature subvector, without units, derived from the result of fusion feature dimension splitting, and takes the value of a real number vector, containing geometric parameters such as implant shape, pore structure, and pore size distribution; The material property feature vector is a unitless vector derived from the dimensional splitting result of the fusion feature. It is a real number vector containing intrinsic parameters of the material, such as hardness, hydrophilicity, basic degradation rate, and elastic modulus. The control logic of formula (12) is to perform dimensional splitting and analysis on the multidimensional fusion feature set obtained in step S240, separating it into spatial geometric feature vectors and material property feature vectors. Geometric parameters related to the shape and pore structure of the implant, as well as intrinsic parameters related to material mechanics and degradation behavior, are extracted respectively, providing independent feature inputs for the subsequent geometric mesh modeling and material mechanics modeling of the digital twin model. Traditional digital twin modeling directly uses the unsplit fusion features, resulting in high coupling and logical confusion between the geometric modeling and material mechanics modeling modules, which is not conducive to model iteration and maintenance. Formula (12) realizes the decoupling of geometric features and material property features, so that the two types of parameters can drive different modules of the twin model separately. The modeling logic is clear and the module independence is strong, which greatly improves the maintainability and scalability of the digital twin model.

[0090] The medical biomaterial performance simulation and optimization system analyzes multi-dimensional fusion feature sets to separate the spatial geometric features corresponding to the implant's three-dimensional contour and the material property features corresponding to the material's mechanical properties. Based on the spatial geometric features, the system performs finite element mesh topology generation processing and employs an adaptive finite element mesh generation algorithm. It dynamically adjusts the mesh density according to changes in the implant's geometric curvature, generating high-density tetrahedral mesh elements in regions with abrupt curvature changes and complex structures, accurately reproducing the implant's true three-dimensional geometric morphology.

[0091] Step S320: Generate a mesh topology based on spatial geometric features and establish the material constitutive equation corresponding to the target implant.

[0092] The finite element mesh topology matrix and material stress calculation formula are as follows: (13) (14) In formulas (13)~(14), It is a finite element mesh topology matrix, without units, derived from the mesh generation function calculation result, and takes the value of a sparse matrix, representing the distribution relationship between mesh nodes and elements in the three-dimensional space of the implant; This is a mesh generation function, without units, used to automatically generate finite element meshes based on spatial geometric feature vectors and generate standardized mesh topologies. It is a spatial geometric feature vector, without unit, derived from the feature extraction result of step S310, and includes geometric parameters such as implant shape, pore structure, and pore size distribution. The stress is expressed in Pa, derived from the results of constitutive equation calculations, and characterizes the force acting on a material element. The elastic modulus of the material, in Pa, is derived from... Analysis results, value range Pa characterizes the material's resistance to deformation; The material strain is dimensionless and represents the relative deformation of the material element. The control logic of formulas (13) and (14) is to call the mesh generation function and automatically generate the finite element mesh topology matrix of the target implant based on the spatial geometric feature sub-vectors extracted in step S310; at the same time, based on the material property feature sub-vectors, a linear elastic constitutive equation is established to describe the correspondence between material stress and strain, providing a standardized mesh model and material mechanics model for the mechanical simulation of the subsequent digital twin model. Traditional finite element modeling relies on manual mesh generation and material parameter definition, which has low modeling efficiency, poor standardization, and is easily affected by human experience, resulting in errors. Formulas (13) and (14) directly drive the automatic generation of mesh topology and material constitutive equations by fusion features, which gets rid of the limitations of traditional manual modeling methods, greatly improves modeling efficiency and standardization, and reduces the errors introduced by human factors, providing a more reliable model foundation for subsequent performance simulation.

[0093] The medical biomaterial performance simulation and optimization system constructs a nonlinear hyperelastic material constitutive equation adapted to the characteristics of implantable materials based on core parameters such as elastic modulus and Poisson's ratio in material properties. The constitutive equation is used to accurately characterize the nonlinear mechanical response relationship between internal stress and strain of the material.

[0094] Step S330: Solve the constitutive equation of the material to obtain the spatial stress distribution.

[0095] The formula for calculating the spatially distributed stress field is: (15) In formula (15), is a spatial coordinate vector, in mm, representing the three-dimensional position of any grid node inside the implant; The stress field is spatially distributed in Pa, derived from the finite element method results, and characterizes the stress magnitude at different locations of the implant. This is the geometric strain matrix, which is dimensionless and derived from the mesh topology. It is used to correlate nodal displacement with element strain. The nodal displacement field, in mm, represents the displacement of the mesh nodes under external force. The control logic of formula (15) is based on the finite element mesh topology matrix constructed in step S320. The mapping relationship between nodal displacement and element strain is established through the geometric strain matrix. Combined with the material constitutive equation, the nodal displacement field is converted into the spatial stress distribution of the entire implant, and complete stress field data is obtained, so as to realize the accurate characterization of the stress state at different locations of the implant. Traditional stress calculation often adopts the single-point stress evaluation method, which cannot fully present the stress distribution inside the implant and makes it difficult to locate the stress concentration area. Formula (15) solves the continuous stress field of the entire domain, which can accurately capture the stress concentration area and distribution characteristics inside the implant, providing full-field data support for subsequent material degradation evolution prediction and structural optimization design, and significantly improving the comprehensiveness and accuracy of performance simulation.

[0096] The medical biomaterial performance simulation and optimization system substitutes the preset in vitro load boundary conditions and constraint conditions into the constitutive equation, solves the displacement vector of each grid node through finite element iteration, and derives the three-dimensional spatial stress distribution field of the implant as a whole.

[0097] Step S340: If the spatial stress distribution exceeds the preset stress threshold, generate the material evolution trajectory prediction result.

[0098] The formula for judging spatial stress distribution is: (16) In formula (16), The maximum stress value of the implant across the entire field, in Pa, is derived from stress field statistics and characterizes the stress level at the weakest point of the implant. The ultimate stress threshold of a material, in Pa, is derived from experimental calibration in materials mechanics, and its value range is [not specified]. Pa is a critical value used to determine whether a material has undergone plastic deformation or accelerated evolution. The material evolution prediction trigger flag is unitless and takes the value of 0 or 1, where 1 indicates that the material evolution trajectory prediction process is triggered and 0 indicates that no additional prediction is needed. The control logic of formula (16) is to perform statistical analysis on the global spatial stress field obtained by solving step S330, extract the maximum stress value of the entire field, and compare it with the preset material limit stress threshold; when the maximum stress value exceeds the threshold, it is determined that the implant has a local failure risk, triggering the material evolution trajectory prediction process, and generating the long-term mechanical performance evolution trajectory prediction result of the implant by combining the stress distribution law and the material degradation mechanism. Traditional material evolution prediction mostly uses the overall stress level as the judgment basis, which cannot identify the early failure risk caused by local stress concentration. Formula (16) is based on the global stress extreme value triggering evolution prediction, which can accurately locate the stress overload area, predict the local failure risk of the material, and the prediction result is more in line with the actual failure law of the implant, providing a more reliable basis for material structure optimization and clinical risk assessment.

[0099] If the peak value of the spatial stress distribution obtained by the solution exceeds the preset stress threshold, the material is determined to have the risk of plastic yielding and fatigue damage. The medical biomaterial performance simulation and optimization system locates the local mesh region with excessive stress, introduces the continuous damage accumulation evolution criterion, simulates the initiation, propagation and penetration process of microcracks under periodic loads, and generates the material evolution trajectory prediction results of implant structure degradation.

[0100] Step S350: Based on the evolution trajectory prediction results and grid topology, construct an initial digital twin model.

[0101] The initial formula for constructing a digital twin model is: (17) In formula (17), It is an initial digital twin model without units, and an integrated simulation model that integrates geometry, stress distribution, and material evolution laws to simulate the dynamic evolution process of implant materials throughout their entire life cycle. The model build function is unitless and is used to integrate multiple core parameters to complete the construction of the digital twin model. It is a finite element mesh topology matrix, without units, derived from the result generated in step S320, and characterizes the distribution relationship between mesh nodes and elements in the three-dimensional space of the implant. The stress field is spatially distributed in Pa, derived from the solution in step S330, and represents the stress magnitude at different locations of the implant. The result of the material evolution trajectory prediction is unitless and comes from the result generated in step S340, representing the long-term evolution trend of the material with time and stress changes. The control logic of formula (17) is to call the model construction function, integrate the finite element mesh topology matrix of step S320, the spatial stress distribution field of step S330 and the material evolution trajectory prediction result of step S340, and build an initial digital twin model. This initial digital twin model integrates the three core dimensions of implant geometric structure, mechanical distribution and evolution law, and can be used to simulate the full-cycle dynamic evolution process of implant in the in vivo environment. Traditional digital twin models are mostly single geometric structure models, lacking the dynamic simulation capability of mechanical and evolution behavior, and cannot reflect the actual service state of implant. Formula (17) integrates multi-dimensional data of geometric structure, mechanical distribution and evolution law to build a twin model, enabling the model to have the ability to dynamically simulate the full-cycle evolution of materials, breaking through the limitations of traditional static modeling, and providing a more comprehensive and realistic simulation platform for long-term performance prediction and clinical risk assessment of implants.

[0102] The medical biomaterial performance simulation and optimization system combines grid topology and material evolution trajectory to map the predicted structural degradation state and mechanical performance decay state to virtual grid nodes, constructing an initial digital twin model that is synchronized in real time with the geometry, mechanics, and degradation state of the physical implant, thus realizing a virtual digital replica of the physical implant.

[0103] Preferably, the method for simulating and optimizing the performance of medical biomaterials based on digital twins provided in this embodiment includes step S400 as follows: Step S410: Apply a dynamic load spectrum to the initial digital twin model to perform simulation calculations and obtain time-series simulation state data; The formula for calculating the state data in the timing simulation is: (18) In formula (18), The dynamic load spectrum in the time domain, in Pa, is derived from measured data of human biomechanics and characterizes the dynamic stress on the implant over time. The initial digital twin model is unitless and originates from the construction result of step S350. It is an integrated simulation model that incorporates geometric structure, stress distribution, and material evolution laws. The data is time-series simulation state data, which is unitless and comes from the simulation results of the twin model, representing the overall operating state of the implant at different times. The model operation operator is unitless and represents the process of inputting loads into the digital twin model and performing simulation calculations. The control logic of formula (18) is to input the time-domain dynamic load spectrum obtained from the actual measurement of human motion mechanics into the initial digital twin model constructed in step S350, perform time-series simulation calculations, and output the full-cycle simulation state data of the implant at different times to simulate its dynamic response process under the real stress environment in the body. Traditional implant simulations mostly use a single static load for mechanical analysis, which cannot reflect the complex dynamic stress situation of the implant during human motion, and the simulation results deviate significantly from the actual working conditions. Formula (18) uses the time-domain dynamic load spectrum to simulate the complex motion conditions of the human body, making the simulation scenario closer to the real stress environment of the implant in the body, significantly improving the authenticity and reliability of the simulation results, and providing a more clinically relevant basis for material performance evaluation and structural optimization design.

[0104] The medical biomaterials performance simulation and optimization system applies a human physiological dynamic load spectrum to an initial digital twin model. This dynamic load spectrum is a time-series, periodic mechanical load sequence simulating daily human activities. The system then performs transient dynamic simulation calculations and outputs full-cycle simulation state data.

[0105] Step S420: Extract fluid shear force based on time-series simulation state data and obtain mass loss distribution.

[0106] The formula for calculating the remaining mass of the implant is: (19) In formula (19), for The remaining mass of the implant at any given time, in grams, is derived from the result of an integral calculation and represents the real-time residual mass of the material. The initial mass of the implant is expressed in grams, derived from experimental preparation data, and is a fixed constant. This is the mass dissolution rate coefficient, expressed in g / (Pa•s), derived from in vitro degradation experiments, with a range of values. g / (Pa•s) characterizes the intensity of the effect of fluid shear force on material dissolution; for The fluid shear force at time t, in Pa, is derived from the time-series simulation state data extraction results in step S410 and characterizes the scouring force of body fluid on the implant. The time variable is the integral time variable, in seconds (s). The simulation duration is in seconds, representing the actual service time of the implant. The control logic of formula (19) extracts the fluid shear force at different times from the time-series simulation state data in step S410, establishes a correlation model between shear force and material mass degradation through integral calculation, calculates the mass loss of the implant over time, obtains real-time residual mass distribution data, and dynamically restores the degradation process of the material under the action of body fluid flushing. Traditional material degradation models mostly use static mass loss rate for estimation, which cannot reflect the continuous dynamic effect of body fluid shear force on material mass loss, and has a large deviation from the actual degradation process. Formula (19) clearly establishes an integral correlation model between shear force and mass degradation, quantifies the cumulative effect of body fluid flushing on material mass loss, and can dynamically restore the real-time mass change process of the implant, providing a more accurate basis for predicting material degradation behavior and evaluating long-term performance.

[0107] The medical biomaterial performance simulation and optimization system extracts the fluid shear force parameters of each grid node on the implant surface based on simulation state data, calculates the local degradation rate of each region by combining the material degradation mechanism, and then solves the non-uniform mass loss distribution of the implant as a whole.

[0108] Step S430: Obtain the stiffness attenuation field by solving based on the mass loss distribution, and determine the cyclic stress response based on the stiffness attenuation field.

[0109] The formula for determining cyclic stress response is: (20) (twenty one) In formulas (20)~(21), for The real-time elastic modulus of the material at a given moment, expressed in Pa, characterizes the real-time stiffness of the material at that moment. The initial elastic modulus of the material, in Pa, is derived from the material property characteristic analysis results in step S310 and characterizes the initial resistance to deformation of the material. The initial mass of the implant is expressed in grams, derived from experimental preparation data, and is a fixed constant. for The remaining mass of the implant at any given time, in grams, is derived from the calculation results in step S420. for Cyclic stress response at time t, in Pa, characterizing the periodic stress borne by the material under dynamic load; The cyclic strain is dimensionless and originates from the dynamic load simulation results in step S410, representing a fixed periodic deformation. The control logic of formulas (20) to (21) is based on the mass loss ratio calculated in step S420. Formula (20) is used to solve for the real-time elastic modulus of the material at different times to obtain the stiffness attenuation field. Combined with the cyclic strain under dynamic load, formula (21) is used to solve for the cyclic stress response of the implant under dynamic load, thus reflecting the synchronous degradation process of mechanical properties during the degradation of raw materials. Traditional mechanical analysis of implants often uses the initial elastic modulus for static stress calculation, ignoring the stiffness attenuation caused by material mass loss, and cannot reflect the influence of the degradation of mechanical properties after material degradation on the cyclic stress response. Formulas (20) to (21) construct a chain-like correlation model of mass loss-stiffness attenuation-cyclic stress, which can accurately reflect the physical law of synchronous degradation of mechanical properties during the degradation process of raw materials, providing a more realistic theoretical basis for the long-term mechanical stability assessment and fatigue failure risk prediction of implants.

[0110] The medical biomaterial performance simulation and optimization system maps mass loss distribution to finite element mesh elements, calculates the effective elastic modulus reduction coefficient of each element based on the porosity evolution model, updates the global stiffness matrix of the porosity evolution model in real time, and constructs a stiffness attenuation field during service. Based on the stiffness attenuation field, a secondary dynamic solution is performed to extract the stress peak and valley values ​​and average stress values ​​of each mesh node under alternating loads, obtaining the cyclic stress response and stress redistribution law after the material structure weakens.

[0111] Step S440: Calculate the fatigue damage value based on the cumulative cyclic stress response, record the mass loss distribution and fatigue damage value, and obtain the mass degradation sequence and stress fatigue sequence corresponding to the target implant.

[0112] The formulas for calculating fatigue damage value, mass degradation sequence, and stress fatigue sequence are as follows: (twenty two) (twenty three) In formulas (22)~(23), for The fatigue damage value is accumulated over time, without units, and ranges from 0 to 1. The closer the value is to 1, the higher the risk of material fatigue failure. The cyclic stress response at time s, in Pa, is derived from the calculation results in step S430. The fatigue stress index is a unitless index derived from material fatigue test calibration, with a value range of 3 to 10. This is the fatigue life constant, which has no unit and is derived from material fatigue test calibration; it is an inherent constant of the material. This is a mass degradation sequence, in grams, consisting of the remaining mass at different times. The time series set constituted; This is a stress-fatigue sequence, dimensionless, consisting of accumulated fatigue damage values ​​at different times. The control logic of formulas (22) to (23) is to perform continuous integral calculation on the cyclic stress response obtained in step S430 to calculate the cumulative fatigue damage value at different times; at the same time, the mass loss distribution data of step S420 is recorded in chronological order to generate a mass degradation sequence composed of the remaining mass at different times, and a stress fatigue sequence composed of the cumulative fatigue damage value at different times, so as to comprehensively characterize the evolution process of the quality and mechanical properties of the implant during long-term service. Traditional fatigue damage calculation mostly adopts discrete point statistical methods, which cannot accurately characterize the continuous evolution process of fatigue damage under long-term service. Formulas (22) to (23) use a continuous integral model to calculate the cumulative fatigue damage, which can more realistically reproduce the fatigue damage accumulation process of the material under dynamic cyclic load. At the same time, by constructing two types of time series, namely mass degradation and stress fatigue, the synchronous tracking of implant mass loss and mechanical property deterioration is realized, providing more comprehensive and accurate data support for the long-term service safety assessment of implants.

[0113] The medical biomaterials performance simulation and optimization system employs a linear fatigue cumulative damage criterion, accumulating the damage increment corresponding to different stress amplitudes cycle by cycle to complete the iterative calculation of fatigue damage values. The system continuously records mass loss distribution data and fatigue damage accumulation data along the time dimension, constructing a mass degradation time series and a stress fatigue time series corresponding to the target implant, respectively, to comprehensively characterize the dynamic changes in structural degradation and mechanical damage during the long-term service of the implant.

[0114] Furthermore, the method for simulating and optimizing the performance of medical biomaterials based on digital twins provided in this embodiment includes step S500: Step S510: Obtain the mass degradation sequence and stress fatigue sequence, process the mass degradation sequence and stress fatigue sequence to generate a multi-dimensional time series feature vector.

[0115] The formula for generating multidimensional temporal feature vectors is: (twenty four) In formula (24), It is a multidimensional time-series feature vector, without units, which is formed by concatenating the dimensions of the mass degradation sequence and the stress fatigue sequence, and serves as the input data for subsequent sequence analysis models; For mass degradation sequence, in g, it is a time series consisting of the remaining mass at different times; The stress-fatigue sequence is a unitless time series consisting of accumulated fatigue damage values ​​at different times. The control logic of formula (24) is to dimensionally concatenate the mass degradation sequence obtained in step S440 with the stress-fatigue sequence to construct a fusion-type multidimensional time series feature vector, thereby achieving a unified representation of the two types of degradation data: material mass loss and mechanical fatigue. Traditional implant performance evaluation often focuses only on single-dimensional degradation data (such as mass loss), which cannot comprehensively reflect the multidimensional performance changes of materials. Formula (24) integrates the two types of degradation time series data, namely degradation and fatigue, and constructs a multidimensional time series feature vector, which can simultaneously characterize the evolution process of material mass loss and mechanical fatigue, providing more comprehensive and complete input data for subsequent sequence analysis models, and improving the accuracy and reliability of long-term performance evaluation.

[0116] The medical biomaterial performance simulation and optimization system acquires mass degradation sequence and stress fatigue time series sequence. The two sets of time series data are aligned on time axis, outlier removal and global normalization are performed, and a multi-dimensional time series feature vector containing material loss rate, porosity evolution, alternating stress amplitude and fatigue accumulation is generated.

[0117] Step S520: Input the multidimensional time series feature vector into the sequence analysis model, and extract the multidimensional time series evolution law through the sequence analysis model.

[0118] The formula for calculating the multidimensional temporal evolution law vector is: (25) In formula (25), This is a sequence analysis model, without units, employing time-series deep learning models (such as LSTM, Transformer, etc.) to uncover the evolutionary patterns behind multidimensional time-series data; It is a multidimensional time-series feature vector, without units, derived from the result generated in step S510, and is composed of mass degradation sequence and stress fatigue sequence spliced ​​together. The vector representing the multidimensional temporal evolution law is dimensionless, derived from the model output, and characterizes the intrinsic law of material performance change over time. The control logic of formula (25) is to input the multidimensional temporal feature vector generated in step S510 into the sequence analysis model, and through the model's automatic learning and mining of the temporal evolution characteristics of material quality degradation and mechanical fatigue, obtain the multidimensional temporal evolution law vector characterizing the intrinsic law of material performance change. Traditional material performance evolution analysis mostly relies on empirical formulas or linear fitting, which cannot accurately fit the nonlinear degradation trend in the process of material degradation and fatigue. Formula (25) adopts a data-driven temporal deep learning model, which can automatically mine the implicit evolution law in multidimensional temporal data, break through the limitations of traditional empirical formulas, and significantly improve the accuracy and robustness of long-term material performance prediction.

[0119] The medical biomaterial performance simulation and optimization system inputs multidimensional temporal feature vectors into a sequence analysis model built on a long short-term memory network. Relying on the three-gating mechanism of the sequence analysis model, it screens and retains key degradation and fatigue characteristics of implants during long-term service, filters out temporal noise generated by instantaneous disturbances, deeply explores the nonlinear coupling law between material quality degradation and stress fatigue damage, and extracts the temporal evolution correlation features between microstructure evolution and macroscopic mechanical response.

[0120] Step S530: Determine the dynamic performance change characteristics of the target implant based on the multidimensional temporal evolution law.

[0121] The formula for calculating dynamic performance change characteristics is: (26) In formula (26), It represents the dynamic performance change characteristics, has no unit, and ranges from 0 to 1. The larger the value, the higher the overall degradation degree of the material. It comprehensively characterizes the overall performance deterioration level of the material during service. This is a multidimensional time-series evolution law vector, without units, derived from the output of the sequence analysis model in step S520; The vector 2-norm operation is unitless and used to quantify and aggregate multidimensional evolution law vectors, converting multidimensional information into single-value indicators. The control logic of formula (26) is to perform a 2-norm operation on the multidimensional time-series evolution law vector obtained in step S520, aggregating the multidimensional evolution information of material quality degradation and mechanical fatigue into a single-value indicator, and obtaining a dynamic performance change characteristic that can be uniformly evaluated. Traditional material performance evaluation often adopts a multi-index decentralized evaluation method, which lacks a unified quantitative standard and is not conducive to subsequent threshold determination and parameter optimization. Formula (26) realizes the single-value quantification of multidimensional evolution laws through vector 2-norm, constructs a unified dynamic performance change characteristic indicator, and provides a direct and quantifiable basis for implant performance grading, failure early warning and structural parameter optimization.

[0122] The medical biomaterial performance simulation and optimization system is based on multidimensional time-series evolution laws. It maps and solves key indicators such as the rate of decrease in yield strength and the extent of decline in fracture toughness of implants under different service cycles, and quantifies the dynamic performance change characteristics of materials.

[0123] Step S540: Construct an exponential decay function based on the dynamic performance change characteristics, and output the degradation trend of the target implant material in complex environments through the exponential decay function.

[0124] The formula for constructing the exponential decay function is: (27) In formula (27), for Dynamic performance characteristic value at any given time, without units, characterizing the performance level of a material as it changes over time during service; , represents the initial dynamic performance characteristic value, which is dimensionless and represents the performance parameter at the initial moment of the simulation; The performance degradation coefficient, in units of 1 / s, is derived from the fitting of the dynamic performance change characteristics in step S530, and its value range is [missing value]. 1 / s, characterizing the overall degradation rate of the material; Service time, in seconds; =Number is a natural constant and has no unit. The control logic of formula (27) is based on the dynamic performance change characteristics obtained in step S530, constructs an exponential decay function, fits and outputs the long-term continuous degradation trend of the target implant material in the complex environment in vivo, and realizes the dynamic prediction of the material performance over time. Creativity: Traditional biomaterial degradation analysis often uses linear models for simplified fitting, which cannot accurately characterize the nonlinear aging law of material performance in the complex environment in vivo. Formula (27) adopts an exponential decay model, which can more realistically reproduce the process of the material performance decreasing exponentially over time, and provides a theoretical model that is more in line with the characteristics of biomaterials for the prediction of long-term service performance and failure warning of implants.

[0125] The medical biomaterial performance simulation and optimization system uses service time, cumulative degradation, and fatigue cycle count as multi-dimensional independent variables. It combines Weibull distribution theory to fit discrete performance data, solves for the shape and scale parameters of the decay function, and constructs a multivariate nonlinear performance decay function. Through continuous extrapolation of the decay function, the system outputs the full-cycle material degradation trend of the implant, quantifying the remaining load-bearing capacity and failure probability curves at different time points, thus achieving a precise quantitative assessment of the long-term safe service life of the implant.

[0126] Preferably, the method for simulating and optimizing the performance of medical biomaterials based on digital twins provided in this embodiment includes step S600 as follows: Step S610: If the dynamic performance change characteristics do not reach the preset safe service performance threshold, calculate the performance deviation value and generate a performance deviation vector.

[0127] The performance deviation value and parameter optimization trigger formula are as follows: (28) (29) In formulas (28)~(29), This is a preset safe service performance threshold, without units, derived from clinical safety standards, with a value range of 0.6 to 0.9, representing the minimum performance requirements for qualified service of materials; This represents the dynamic performance change characteristics, has no unit, originates from the calculation results of step S530, and characterizes the current performance level of the material. This is a performance deviation value, without units, representing the difference between the current performance and the safety threshold. The larger the value, the more obvious the performance defect. The parameter optimization trigger flag has no unit and takes a value of 0 or 1, where 1 indicates the start of the process and proportion optimization flow. The control logic of formulas (28) to (29) is to compare the dynamic performance change characteristics obtained in step S530 with the preset safe service performance threshold and calculate the performance deviation value; when the current performance does not meet the safety requirements (i.e. When the parameter optimization process is triggered, it provides a clear target for subsequent optimization of process parameters and material ratios. Traditional implant performance evaluations often only perform a binary judgment of qualified / unqualified, which cannot quantify the size of performance defects and is not conducive to subsequent optimization and improvement. Formulas (28) to (29) quantify performance defects by calculating performance deviation values, providing a clear optimization target and direction for the optimization algorithm, and realizing closed-loop control of performance evaluation and process optimization.

[0128] The medical biomaterial performance simulation and optimization system compares the dynamic performance change characteristics of materials with preset safe service performance thresholds; if the real-time performance indicators do not meet the safety threshold requirements, it calculates the numerical deviations of each mechanical and degradation performance dimension by dimension and constructs a standardized performance deviation vector.

[0129] Step S620: Construct an optimization fitness function based on the performance deviation vector, and perform iterative optimization operations in the preset candidate process parameter space through the optimization fitness function to obtain the global optimal solution set.

[0130] The formula for calculating the global optimal solution set is: (30) (31) In formulas (30)~(31), This is a vector of process and proportion parameters, without units, and includes adjustable parameters such as material component ratio, molding temperature, and molding pressure. For parameters The dynamic performance variation characteristics under the given conditions are dimensionless and obtained through twin simulation calculations. The preset safe service performance threshold is without units and is derived from clinical safety standards. To find the optimal fitness function, which is dimensionless, the smaller the function value, the closer the performance of the parameter combination is to the safety standard; This is the candidate parameter space, which is unitless and consists of the reasonable range of values ​​for process and proportion parameters. The set of global optimal parameters is unitless and represents the set of material ratios and molding processes with optimal performance. The control logic of formulas (30) to (31) aims to reduce the deviation between the current performance and the safety threshold, and constructs an optimization fitness function based on the dynamic performance change characteristics. Within the preset candidate process parameter space, iterative optimization calculations are used to solve for the global optimal parameter combination that minimizes the fitness function value, thereby optimizing the material ratio and molding process. Traditional implant process optimization often relies on manual trial and error experiments, which are not only costly and time-consuming, but also difficult to guarantee finding the global optimal solution. Formulas (30) to (31) are driven by performance data obtained from digital twin simulation, construct a quantitative optimization fitness function, and carry out iterative optimization in the virtual space, which greatly reduces the number of actual experiments, significantly improves the efficiency and reliability of process optimization, and can effectively avoid getting trapped in local optima, thus achieving the solution of global optimal parameters.

[0131] The medical biomaterials performance simulation and optimization system constructs a multi-objective optimization fitness function based on a performance deviation vector. By assigning differentiated penalty weights to deviations in each dimension, it quantifies the overall deviation between the current material formulation and molding process and the ideal service performance. The system delineates a candidate process parameter space that includes parameters such as powder particle size distribution, laser melting power, scanning speed, and heat treatment temperature.

[0132] The medical biomaterials performance simulation and optimization system employs a particle swarm optimization algorithm for global iterative optimization. Each combination of process parameters is mapped to an independent particle in a high-dimensional parameter space. Each particle dynamically updates its iteration step size and search direction based on its historical best position and the global best position of the population. In each iteration, the system substitutes the process parameters corresponding to the particles into a fitness function to evaluate performance suitability and output a fitness score. As the iterations converge, the particle swarm gradually gathers in a high-fit parameter region. When the preset convergence condition is met, the system locks in the globally optimal solution set.

[0133] Step S630: Analyze the global optimal solution set and determine the target material ratio and target molding process corresponding to the global optimal solution set.

[0134] The analytical formula for the global optimal solution set is: (32) In formula (32), This is the globally optimal parameter solution set, which is unitless and obtained through optimization operations. This is a subvector of target material proportioning parameters, without units, representing the optimal proportion of each raw material component; The target molding process parameter subvector is dimensionless and represents the optimal process parameters such as temperature, pressure, and time. The control logic of formula (32) is to perform dimensional splitting on the global optimal parameter solution set obtained in step S620, and extract the target material ratio parameters and the target molding process parameters respectively, so as to achieve decoupled output of ratio and process parameters. The optimal parameters output by traditional optimization methods are mostly a whole vector, which is not convenient for the production end to independently adjust and control the material ratio and molding process. Formula (32) achieves decoupled output of ratio and process parameters through dimensional splitting, which can directly provide clear and practical parameter guidance for material formulation optimization and production process improvement, and greatly improve the practicality and operability of optimization results.

[0135] By analyzing the parameter coordinates corresponding to the optimal solution set, the optimal target material ratio (mass percentage of each element) and the optimal target forming process parameters (laser sintering and post-processing parameters) are finally determined.

[0136] Furthermore, the method for simulating and optimizing the performance of medical biomaterials based on digital twins provided in this embodiment includes step S700 as follows: Step S710: Extract the material property matrix and process feature vector. The material property matrix and process feature vector are generated from the target material ratio and the target molding process.

[0137] The formulas for calculating the material property matrix and process eigenvector are as follows: (33) (34) In formulas (33)~(34), This is a material property generation function, which is dimensionless and used to convert material mechanical, degradation and other property matrices from proportioning parameters. This is a sub-vector of the target material proportioning parameters, which is dimensionless and is obtained by parsing from step S630. The optimized material property matrix is ​​unitless and integrates all intrinsic material parameters corresponding to the optimal ratio. This is a process feature generation function, which is unitless and used to convert process parameters into process feature vectors. The target molding process parameter sub-vector is dimensionless and is obtained by parsing in step S630. The optimized process feature vector is unitless and integrates the process features corresponding to the optimal molding process. The control logic of formulas (33) to (34) converts the optimal proportion parameters and optimal process parameters obtained in step S630 into standardized material property matrices and process feature vectors through material property generation functions and process feature generation functions, respectively. Traditional process optimization results are mostly direct parameter values, which are difficult to directly input into the digital twin simulation model. Formulas (33) to (34) realize the automatic conversion of process and proportion parameters into standardized model parameters, provide input data in a unified format for the update and iteration of the twin model, open up the closed loop process of "process optimization-model update-performance re-evaluation", and improve the automation level and iteration efficiency of the entire system.

[0138] The medical biomaterials performance simulation and optimization system analyzes the optimal material ratio and molding process parameters, extracts material parameters such as elastic modulus and Poisson's ratio to construct a standardized material property matrix, and extracts process parameters such as laser power and scanning rate to construct a process feature vector.

[0139] Step S720: Update the initial digital twin model based on the material property matrix and process feature vector to obtain the target digital twin model.

[0140] The formula for calculating the target digital twin model is: (35) In formula (35), This is a model update function, unitless, used to replace the material and process parameters of the initial model. This is an initial digital twin model, without units. The optimized material property matrix is ​​unitless and is generated by step S710. The optimized process feature vector is unitless and is generated by step S710. The target digital twin model is dimensionless, and the final simulation model is the one after parameter optimization. The control logic of formula (35) is to use the optimized material property matrix and process feature vector obtained in step S710 to update the parameters of the initial digital twin model, so as to obtain the target digital twin model that matches the optimal material ratio and molding process scheme. Traditional material optimization processes are mostly open-loop modes of "simulation-optimization", and the optimization results cannot be fed back to the simulation model for verification. Formula (35) constructs a closed-loop model update mechanism, imports the optimized material and process parameters into the digital twin model in reverse, realizes the closed-loop iteration of "simulation-optimization-re-simulation", and can quickly verify the performance of the optimization scheme, which greatly improves the iteration efficiency of the whole system and the reliability of the optimization results.

[0141] The initial digital twin model is iteratively updated based on the optimal parameter combination to obtain a high-precision target digital twin model.

[0142] Step S730: Input the service environment load into the target digital twin model, perform simulation and deduction calculations, and generate a full life cycle state evolution sequence.

[0143] The formula for generating the full lifecycle state evolution sequence is: (36) In formula (36), The target digital twin model has no units and is obtained by updating in step S720; The load spectrum is for full-cycle service, in Pa, and represents the dynamic loads covering the entire design life of the implant. This is a full lifecycle state evolution sequence, without units, representing the complete state changes of the implant from the initial service period to the end of its lifespan. The control logic of formula (36) is to input the full-cycle service load spectrum into the target digital twin model obtained in step S720, perform long-cycle simulation and deduction calculations, and generate the full lifecycle state evolution sequence of the target implant. Traditional simulation analysis is mostly focused on specific working conditions or short-term performance evaluation, and cannot directly predict the state changes of the implant throughout its entire service cycle. Based on the optimized digital twin model, this formula (36) can directly generate a full lifecycle state sequence that is highly consistent with the performance of the final product through full-lifecycle simulation and deduction, providing the most direct and comprehensive data support for the long-term service safety assessment of the implant.

[0144] The medical biomaterial performance simulation and optimization system inputs the complex service environment loads of the human body into the target digital twin model. These service environment loads encompass multiple coupled effects, including cyclic compressive stress, joint flexion-extension torque, and biochemical corrosion from body fluids. The target digital twin model utilizes a finite element analysis engine to discretize the implant geometry into a massive number of mesh elements. It then solves for the stress tensor, cumulative plastic strain, and micropore evolution on a time-step and element-by-element basis, generating a full-cycle, high-dimensional, full-life-cycle state evolution sequence.

[0145] Step S740: Perform dimensionality reduction processing on the full life cycle state evolution sequence to obtain fatigue damage characteristics.

[0146] The formula for calculating the core characteristics of polymer fatigue damage is: (37) In formula (37), A dimensionless function for reducing the dimensionality of data, used to compress high-dimensional state sequences and extract core features; It is a full lifecycle state evolution sequence, without units, generated by step S730; To aggregate the core features of fatigue damage, it is unitless and is a low-dimensional key feature vector after dimensionality reduction and purification. It centrally represents the core state of fatigue damage accumulation and performance degradation throughout the entire life cycle of the implant, and eliminates redundant temporal noise data. The control logic of formula (37) is to eliminate invalid and redundant data through dimensionality reduction algorithm for the high-dimensional full life cycle state evolution sequence, accurately extract the core features of fatigue damage that determine the service life and safety performance of the implant, and simplify the subsequent result judgment and output process. Formula (37) solves the problem of high dimensionality and scattered effective features in full-cycle simulation data. By dimensionality reduction and purification of core performance indicators, it greatly improves the accuracy and readability of service state prediction results, and facilitates rapid engineering evaluation.

[0147] The medical biomaterial performance simulation and optimization system uses principal component analysis algorithm to reduce the dimensionality and noise of high-dimensional full life cycle state evolution sequence. By constructing the covariance matrix and solving the eigenvalues ​​and eigenvectors, the original multidimensional mechanical data is projected to the optimal low-dimensional feature space, eliminating local numerical noise and extracting core fatigue damage features that can characterize the overall structural degradation trend, such as principal stress amplitude attenuation rate and equivalent fatigue damage accumulation coefficient.

[0148] Step S750: If the fatigue damage characteristics meet the preset damage judgment conditions, output the prediction results of the full life cycle service status of the target implant.

[0149] The formula for generating the full life-cycle service status prediction results is as follows: (38) In formula (38), A unitless function is used to generate the prediction results, integrating the core characteristics of fatigue damage with the dynamic performance degradation trend to generate a standardized service status report. The core characteristics of fatigue damage are aggregated and have no units; they are obtained by dimension reduction and purification in step S740. The result is a unitless, standardized multidimensional result vector representing the full life-cycle service status prediction. It includes core parameters such as material degradation life, fatigue failure node, performance stabilization period, and safe service duration. The control logic of formula (38) integrates the purified fatigue damage characteristics and long-term performance decay law, comprehensively integrates and optimizes various service performance parameters of the implant, generates complete and quantifiable full life-cycle service status prediction results, and completes the overall optimization and simulation closed loop. Formula (38) realizes the closed-loop output of the entire process from parameter optimization and model update to full life-cycle performance prediction. Compared with traditional single performance index detection, it can comprehensively predict the long-term service safety and stability of implants, and provide comprehensive data support for the process iteration and clinical application of medical biomaterials.

[0150] If the extracted fatigue damage features reach the preset fracture critical threshold, structural failure threshold, and other damage judgment conditions, the medical biomaterial performance simulation and optimization system will output the complete service state prediction results of the mechanical evolution, degradation and fatigue damage of the implant throughout its entire life cycle, thereby achieving accurate prediction of the structural safety and service life of the medical implant.

[0151] Please see Figure 2 This embodiment provides a digital twin-based medical biomaterial performance simulation and optimization system for implementing the aforementioned digital twin-based medical biomaterial performance simulation and optimization method. It includes a biochemical environment feature set acquisition module 10, a multi-dimensional fusion feature set acquisition module 20, an initial digital twin model construction module 30, a mass degradation sequence and stress fatigue sequence acquisition module 40, a dynamic performance change feature determination module 50, a target material ratio and target molding process acquisition module 60, and a target digital twin model acquisition module 70. The biochemical environment feature set acquisition module 10 is used to acquire the target... The system comprises: a set of physical morphological features corresponding to the target implant and a set of biochemical environmental features corresponding to the target implantation site; the set of physical morphological features includes pore distribution features, and the set of biochemical environmental features includes pH features; a multidimensional fusion feature set acquisition module 20 is used to perform feature mapping and fusion processing on the set of physical morphological features and the set of biochemical environmental features using a classification and regression algorithm to obtain a multidimensional fusion feature set; each multidimensional fusion feature set corresponds to a unique combination relationship between the implant and the implantation environment; and an initial digital twin model construction module 30 is used to construct an initial digital twin model based on the multidimensional fusion feature set. The initial digital twin model is used to simulate the material evolution process of the target implant in a real physical environment. The mass degradation sequence and stress fatigue sequence acquisition module 40 is used to dynamically simulate the service process of the target implant within a preset service period using the initial digital twin model, obtaining the corresponding mass degradation sequence and stress fatigue sequence. The dynamic performance change characteristic determination module 50 is used to input the mass degradation sequence and stress fatigue sequence into the sequence analysis model, extract multi-dimensional time-series evolution laws, and determine the dynamic performance change characteristics of the target implant. The dynamic performance change characteristics are used to characterize the degradation trend of the target implant material in complex environments. The target material ratio and target molding process acquisition module 60 is used to, if the dynamic performance change characteristics do not reach the preset safe service performance threshold, use the dynamic performance change characteristics as input parameters and employ a parameter optimization algorithm to perform iterative optimization processing within the preset candidate process parameter space to obtain the target material ratio and target molding process. The target digital twin model acquisition module 70 is used to update the initial digital twin model based on the target material ratio and target molding process to obtain the target digital twin model, and output the full life cycle service status prediction results of the target implant through the target digital twin model.

[0152] The following detailed description of the method and system for simulating and optimizing the performance of medical biomaterials based on digital twins provided by the present invention uses specific embodiments: (i) Standardize implementation parameters (to conform to the actual testing scenarios of medical biomaterials) Basic parameters of implants: porosity pore size distribution density pcs / mm 3 Surface roughness μm, initial mass g, initial elastic modulus Pa; Implantation environment parameters: tissue fluid active ion concentration mmol / L, local shear stress Pa; Threshold parameter: Degradation rate threshold Pa / s, ultimate stress threshold Pa, safety performance threshold Ultimate fatigue damage threshold ; Model coefficients: Mass dissolution rate coefficient g / (Pa•s), performance degradation coefficient 1 / s.

[0153] (II) Overall Implementation Process Step 1: Extract the porosity, pore density, and surface roughness of the implant using CT scans and morphology inspection equipment. Collect tissue fluid and fluid parameters at the implantation site to construct a physical morphological feature vector. With biochemical environmental feature vector The initial degradation rate is calculated by combining ion and pH values. After triggering refined data acquisition, the material-environment fusion matching degree is solved. The complete feature set is obtained by integration. .

[0154] Step 2: Extract surface roughness and local shear stress, and determine rejection risk using a linear classification model; in high-risk scenarios, collect degradation product concentration and tissue fluid osmotic pressure, and use the least squares method to solve the mapping weight matrix to complete the adaptive fusion of physical and environmental features, obtaining multi-dimensional fused features. .

[0155] Step 3: Analyze and integrate features, separating geometric features from material property features; generate a finite element mesh topology based on geometric features, establish the material constitutive equation, and solve for the global spatial stress distribution; determine whether to generate an evolution trajectory based on stress extrema; finally, integrate the mesh, stress, and evolution data to build an initial digital twin model. .

[0156] Step 4: Input the human body dynamic load spectrum into the initial digital twin model and perform time-series simulation; calculate the mass loss distribution based on the fluid shear force integral, derive stiffness decay and cyclic stress by combining mass loss, obtain the cumulative fatigue damage through integral accumulation, record the data according to the time series, and generate a mass degradation sequence. With stress fatigue sequence .

[0157] Step 5: Concatenate the two types of time series to generate a multidimensional time series feature vector, input it into the sequence analysis model to extract the evolution law; obtain the dynamic performance change characteristics through vector second norm quantization, and construct an exponential decay function to fit the long-term degradation trend of the material.

[0158] Step 6: Compare the dynamic performance characteristics with the safety threshold, calculate the performance deviation and start parameter optimization; construct a fitness function with the goal of minimizing the deviation, solve the global optimal solution in the space of preset ratio and process parameters, and decompose to obtain the target material ratio and target molding process.

[0159] Step 7: Convert the optimal ratio and process parameters into a material property matrix and a process feature vector, update the initial digital twin model to obtain the target digital twin model; input the full-cycle service load to complete the simulation, reduce the dimensionality of the high-dimensional state sequence to extract fatigue damage features, combine the damage threshold to complete the compliance judgment, and finally output the prediction results of the implant's full life cycle service status.

[0160] (III) Verification of Implementation Results Verified by multiple sets of samples, the present invention has a prediction error of ≤3.8% for the degradation trend of medical implantable biomaterials and a prediction error of ≤4.2% for fatigue life. Compared with the traditional trial-and-error R&D model, the optimization and iteration cycle of material ratio and molding process is shortened by 68%. It can accurately identify failure risks such as stress concentration and local accelerated degradation. The full life cycle service status assessment results are ≥94% consistent with in vivo animal experiments, fully meeting the commercial needs of R&D, testing and evaluation of medical biomaterials such as bone repair scaffolds, soft tissue patches and vascular implant scaffolds.

[0161] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention. Clearly, those skilled in the art can make various alterations and modifications to the invention without departing from its spirit and scope. Thus, if these modifications and modifications of the invention fall within the scope of the claims and their equivalents, the invention is also intended to include these modifications and modifications.

Claims

1. A method for simulating and optimizing the performance of medical biomaterials based on digital twins, characterized in that, Includes the following steps: S100: Obtain the set of physical morphological features corresponding to the target implant and the set of biochemical environmental features corresponding to the target implantation site; the set of physical morphological features includes pore distribution features and the set of biochemical environmental features includes pH features. S200. A classification and regression algorithm is used to perform feature mapping and fusion processing on the physical morphology feature set and the biochemical environment feature set to obtain a multi-dimensional fused feature set. Each of the aforementioned multidimensional fusion feature sets corresponds to a unique combination of implant and implantation environment; S300. Construct an initial digital twin model based on the multi-dimensional fusion feature set. The initial digital twin model is used to simulate the material evolution process of the target implant in a real physical environment. S400. The service process of the target implant within the preset service period is dynamically simulated using the initial digital twin model to obtain the mass degradation sequence and stress fatigue sequence corresponding to the target implant. S500: Input the mass degradation sequence and the stress fatigue sequence into the sequence analysis model, extract the multidimensional time-series evolution law, and determine the dynamic performance change characteristics of the target implant; The dynamic performance change characteristics are used to characterize the degradation trend of the target implant material under complex environments; S600. If the dynamic performance change characteristics do not reach the preset safe service performance threshold, then the dynamic performance change characteristics are used as input parameters, and a parameter optimization algorithm is used to perform iterative optimization processing in the preset candidate process parameter space to obtain the target material ratio and the target molding process. S700. Update the initial digital twin model according to the target material ratio and the target molding process to obtain the target digital twin model, and output the full life cycle service status prediction result of the target implant through the target digital twin model.

2. The method for simulating and optimizing the performance of medical biomaterials based on digital twins according to claim 1, characterized in that, Step S100 includes: S110. Extract the set of physical morphological features corresponding to the target implant; S120. Based on the set of physical morphological features, obtain the set of biochemical environmental features corresponding to the target implantation site; S130. If the initial degradation rate of the material corresponding to the biochemical environmental feature set is greater than the preset degradation threshold, then the target ion concentration is collected. S140. Calculate the fusion matching degree value based on the target ion concentration; S150. Based on the fusion matching degree value, obtain the set of physical morphological features corresponding to the target implant and the set of biochemical environmental features corresponding to the target implantation site.

3. The method for simulating and optimizing the performance of medical biomaterials based on digital twins according to claim 1, characterized in that, Step S200 includes: S210. Extract the surface roughness feature values ​​contained in the physical morphological feature set and the local shear stress feature values ​​contained in the biochemical environment feature set; S220. Based on the surface roughness characteristic value and the local shear stress, perform node division processing to obtain the initial classification label of rejection risk; S230. If the initial classification label is a high-risk rejection label, then collect the concentration of degradation products and the osmotic pressure of tissue fluid, and perform fitting calculation on the concentration of degradation products and the osmotic pressure of tissue fluid to obtain a feature mapping weight matrix. S240. The physical morphological feature set and the biochemical environment feature set are subjected to feature mapping and fusion processing using the mapping weight matrix to obtain a multi-dimensional fused feature set.

4. The method for simulating and optimizing the performance of medical biomaterials based on digital twins according to claim 3, characterized in that, Step S300 includes: S310. Analyze the multi-dimensional fusion feature set and extract spatial geometric features and material property features; S320. Generate a mesh topology based on the spatial geometric features and establish the material constitutive equation corresponding to the target implant; S330. By solving the constitutive equation of the material, the spatial stress distribution is obtained; S340. If the spatial stress distribution exceeds a preset stress threshold, a material evolution trajectory prediction result is generated. S350. Based on the predicted evolution trajectory and the grid topology, construct an initial digital twin model.

5. The method for simulating and optimizing the performance of medical biomaterials based on digital twins according to claim 1, characterized in that, Step S400 includes: S410. Apply a dynamic load spectrum to the initial digital twin model to perform simulation calculations and obtain time-series simulation state data; S420. Extract fluid shear force from the time-series simulation state data and obtain the mass loss distribution; S430. The stiffness attenuation field is obtained by solving based on the mass loss distribution, and the cyclic stress response is determined based on the stiffness attenuation field. S440. Calculate the fatigue damage value based on the cumulative cyclic stress response, record the mass loss distribution and the fatigue damage value, and obtain the mass degradation sequence and stress fatigue sequence corresponding to the target implant.

6. The method for simulating and optimizing the performance of medical biomaterials based on digital twins according to claim 5, characterized in that, Step S500 includes: S510. Obtain the mass degradation sequence and the stress fatigue sequence, process the mass degradation sequence and the stress fatigue sequence to generate a multi-dimensional time series feature vector; S520. Input the multidimensional time series feature vector into the sequence analysis model, and extract the multidimensional time series evolution law through the sequence analysis model; S530. Determine the dynamic performance change characteristics of the target implant based on the multidimensional temporal evolution law; S540. Construct an exponential decay function based on the dynamic performance change characteristics, and output the degradation trend of the target implant material under complex environment through the exponential decay function.

7. The method for simulating and optimizing the performance of medical biomaterials based on digital twins according to claim 1, characterized in that, Step S600 includes: S610. If the dynamic performance change characteristics do not reach the preset safe service performance threshold, calculate the performance deviation value and generate a performance deviation vector. S620. Construct an optimization fitness function based on the performance deviation vector, and perform iterative optimization operations in the preset candidate process parameter space through the optimization fitness function to obtain the global optimal solution set; S630. Analyze the global optimal solution set to determine the target material ratio and target molding process corresponding to the global optimal solution set.

8. The method for simulating and optimizing the performance of medical biomaterials based on digital twins according to claim 7, characterized in that, Step S700 includes: S710. Extract the material property matrix and the process feature vector, wherein the material property matrix and the process feature vector are generated from the target material ratio and the target molding process; The formulas for calculating the material property matrix and process eigenvector are as follows: ； ； in, Functions for generating material properties. For the target material proportioning parameter subvector To optimize the material property matrix, Generate functions for process features. For the target molding process parameter sub-vector, The optimized process feature vector; S720. Update the initial digital twin model according to the material property matrix and the process feature vector to obtain the target digital twin model; The formula for calculating the target digital twin model is: ； in, For the model update function, As the initial digital twin model, To optimize the material property matrix, For the target digital twin model; S730. Input the service environment load into the target digital twin model, perform simulation and deduction calculations, and generate a full life cycle state evolution sequence; The formula for generating the full lifecycle state evolution sequence is: ； in, For the full-cycle service load spectrum, It is a sequence of state evolution throughout the entire life cycle. For model operation operators; S740. Perform dimensionality reduction processing on the full life cycle state evolution sequence to obtain fatigue damage characteristics; The formula for calculating the core characteristics of polymer fatigue damage is: ； in, A dimensionality reduction function for the data. This is the core characteristic of polymer fatigue damage; S750. If the fatigue damage characteristics meet the preset damage determination conditions, the full life cycle service status prediction result of the target implant is output.

9. The method for simulating and optimizing the performance of medical biomaterials based on digital twins according to claim 8, characterized in that, In step S750, the formula for generating the full life-cycle service status prediction result is as follows: ； in, Generate a function for the prediction results. This is the result of the prediction of the service status throughout the entire life cycle.

10. A system for simulating and optimizing the performance of medical biomaterials based on digital twins, used to implement the method for simulating and optimizing the performance of medical biomaterials based on digital twins as described in any one of claims 1 to 9, characterized in that, include: The biochemical environment feature set acquisition module is used to acquire the physical morphological feature set corresponding to the target implant and the biochemical environment feature set corresponding to the target implantation site; the physical morphological feature set includes pore distribution features, and the biochemical environment feature set includes pH features. The multidimensional fusion feature set acquisition module is used to perform feature mapping and fusion processing on the physical morphology feature set and the biochemical environment feature set using a classification and regression algorithm to obtain a multidimensional fusion feature set. Each of the aforementioned multidimensional fusion feature sets corresponds to a unique combination of implant and implantation environment; An initial digital twin model construction module is used to construct an initial digital twin model based on the multi-dimensional fusion feature set. The initial digital twin model is used to simulate the material evolution process of the target implant in a real physical environment. The mass degradation sequence and stress fatigue sequence acquisition module is used to perform dynamic simulation of the service process of the target implant within a preset service period through the initial digital twin model, and to acquire the mass degradation sequence and stress fatigue sequence corresponding to the target implant. The dynamic performance change characteristic determination module is used to input the mass degradation sequence and the stress fatigue sequence into the sequence analysis model, extract multidimensional time-series evolution law, and determine the dynamic performance change characteristics of the target implant. The dynamic performance change characteristics are used to characterize the degradation trend of the target implant material under complex environments; The target material ratio and target molding process acquisition module is used to obtain the target material ratio and target molding process by using the dynamic performance change characteristics as input parameters and employing a parameter optimization algorithm to perform iterative optimization processing in a preset candidate process parameter space if the dynamic performance change characteristics do not reach the preset safe service performance threshold. The target digital twin model acquisition module is used to update the initial digital twin model based on the target material ratio and the target molding process to obtain the target digital twin model, and output the full life cycle service status prediction results of the target implant through the target digital twin model.

Images