Laser additive manufacturing process parameter database establishment method and system and storage medium

Abstract

The invention discloses an establishment method and system of a laser additive manufacturing process parameter database and a storage medium, and belongs to the technical field of additive manufacturing digitization. The method comprises the following steps: firstly, collecting basic physical property data of powder and nickel-based high-temperature alloy and related parameters of laser; commercialized simulation software is used for simulation, and simulation results under the action of various main parameters are obtained. And selecting corresponding parameters to perform an experiment according to a main result of simulation to obtain an experiment result, matching the experiment result obtained under the corresponding parameters with a simulation result, and setting a certain error limit to realize correction. The experiment result and the result obtained by simulation are used as the basis. And then corresponding data is established. A database used by a large model among different nickel-based high-temperature alloys, powder and laser is established according to the method. The method solves the problem that a big data source is not easy to establish when a big model is established. And meanwhile, the method has certain accuracy.

Description

Technical Field

[0001] This invention relates to the field of digital and intelligent additive manufacturing (3D printing) technology, specifically to a method, system and storage medium for establishing a laser additive manufacturing process parameter database, which is particularly applicable to laser powder bed melting and laser directional energy deposition processes for high-performance metal materials such as nickel-based superalloys. Background Technology

[0002] Additive manufacturing, due to its unique manufacturing methods, has been rapidly adopted in fields such as aerospace, medical and transportation, and the mold industry. Among these, laser additive manufacturing is the most widely used. However, compared to traditional casting and welding industries, laser additive manufacturing involves a much larger number of data and variables, thus posing a significant challenge to ensuring the quality of the formed product.

[0003] The parameter system of laser additive manufacturing (taking mainstream laser powder bed fusion (L-PBF) and laser directed energy deposition (L-DED) as examples) directly determines the density, mechanical properties, geometric accuracy, surface quality, and production efficiency of the formed parts. Furthermore, there is a strong coupling relationship between these parameters (e.g., energy density is determined by multiple parameters), meaning that a change in one parameter affects the entire process. Considering the forming requirements of high-temperature alloys and titanium alloys in aerospace and high-end manufacturing fields (such as engine blades and complex structural components), the core parameters and their impacts can be categorized into five dimensions: laser, scanning, powder, environment, and auxiliary processes. All of these directly influence the manufacturing outcome, summarized as follows: Laser-related parameters (core energy input) directly determine the temperature field, fluidity, and solidification rate of the molten pool, and are crucial for controlling defects (porosity, cracks) and microstructure. These parameters mainly include: laser spot size and laser energy distribution pattern (Gaussian beam, flat-top beam, or figure-eight pattern, etc.), laser power, and laser wavelength.

[0004] Scanning strategy parameters (core of forming accuracy and stress control): The scanning strategy determines the interaction path between the laser and the powder, directly affecting the temperature field distribution, residual stress, and grain orientation, and is crucial for complex structures (such as lattice structures and irregular channels). These mainly include scanning path parameters, scanning path (scanning spacing (h), scanning direction), and filling strategy parameters (filling method (straight line / ring / spiral) / cross-shaped method / etc.).

[0005] Powder characteristic parameters (fundamentals of fusion quality): The physicochemical properties of powder determine its uniformity of spreading, flowability, and interaction with the laser, and are key to "incoming material quality." The saying "a skilled cook cannot cook without rice" perfectly illustrates this point. Powder is the raw material for additive manufacturing, especially affecting the forming of refractory materials such as high-temperature alloys and titanium alloys. These parameters include powder geometry (sphericity, particle size and its distribution), powder physical parameters (sphericity, absorbance, flowability (Hall flow rate)), and powder chemical parameters (purity, content of impurities such as oxygen and nitrogen). Process environment parameters (stability assurance) control the atmosphere, temperature, and other conditions during the molding process to prevent material oxidation or performance degradation, which is especially important for active metals (titanium alloys, aluminum alloys) and high-temperature alloys. These parameters include atmosphere control parameters (protective gas type (Ar / N2), oxygen content, and even humidity), substrate-related parameters (substrate temperature, preheating method), and layer parameters (layer thickness, etc.).

[0006] Auxiliary process parameters (performance optimization supplements) are used to improve energy distribution or solidification processes through auxiliary means, and are often used in high-end manufacturing to improve the reliability of parts. Examples include energy compensation parameters (edge ​​scan power correction) and post-processing related parameters (remelting times, heat treatment processes such as annealing and hot isostatic pressing, and other post-processing processes such as shot peening and polishing).

[0007] The parameters of laser additive manufacturing are essentially a closed loop of "energy input - material response - forming quality". In high-end manufacturing, parameters need to be optimized by combining material characteristics (such as the hot cracking sensitivity of high-temperature alloys), part functions (load-bearing / non-load-bearing), and equipment capabilities (laser power range, scanning accuracy).

[0008] High-temperature alloys have a very broad prospect for additive manufacturing applications in aerospace, petrochemicals, and other fields. However, the requirements are also extremely stringent, allowing no defects whatsoever; otherwise, their harsh and complex service environment could lead to aircraft crashes, loss of life, and huge losses in the energy and chemical industries. There are so many influencing parameters, each with a significant impact, and these parameters are interconnected, meaning that a change in one can have far-reaching consequences. Without artificial intelligence, a reliable dataset requires extensive parameter experiments, scale-up experiments, and long-term service accumulation by an individual or organization. This is the fundamental reason why the development and finalization cycle of an engine takes at least ten to fifteen years. Similarly, an experienced engineer needs long-term experience accumulation and exploration to develop a reliable experimental plan and practice. Developing a new process requires a long cycle and huge amounts of funding. If artificial intelligence can be used to assist in the development of a corresponding database—the large model database mentioned in this invention—the development cycle and experimental costs will be greatly reduced and shortened. In practical applications, people usually achieve a balance between performance, efficiency, and cost through orthogonal experiments, response surface methodology (RSM), or AI algorithms (such as machine learning to optimize energy density), which is especially suitable for scenarios with extremely high reliability requirements, such as aerospace. These experiments are all very time-consuming and labor-intensive.

[0009] The data establishment of large-scale models (such as artificial intelligence language models and the additive manufacturing large-scale models mentioned in this invention) mainly involves data preparation, model training, and knowledge management. As the saying goes, "Even a skilled cook cannot make a meal without rice," and one of the fundamental sources of large-scale model establishment and application is data establishment. This is the very foundation. The data sources of large-scale model databases are rich and diverse. Based on dimensions such as the data's universality, specialization, and the entity to which it belongs, they can be divided into three main categories: general public data, vertical professional data, and enterprise-owned data. Each category further includes various specific data formats. The main popular methods for establishing such databases are as follows. To illustrate the innovation of this invention, several traditional methods are briefly described below: 1. General Public Data This type of data is relatively easy to obtain, has a wide coverage, and is large in scale. It is the basic data source for the pre-training stage of large models and can help models master basic language rules and common sense.

[0010] Web page data: This is one of the most crucial sources. Data sets like CommonCrawl, built by a non-profit organization, have been continuously crawling web pages since 2008, accumulating petabytes of data and covering original web pages in multiple languages, metadata, etc. Other examples include Google's C4 dataset and the WanJuan-CC dataset from the Shanghai Artificial Intelligence Laboratory, all of which are high-quality corpora formed after filtering and cleaning web page data. Furthermore, Chinese-language web page datasets such as ChineseWebText and WuDaoCorporaText have also provided a large amount of Chinese corpus for domestic large-scale models.

[0011] Public encyclopedia and book data: Wikipedia is a commonly used data source, covering more than 20 languages. The data is well-organized and is often used by models such as LLaMA and GPT-NEOX. As for books, there is BookCorpus built by the University of Toronto and MIT, as well as a large number of e-book resources provided by projects such as Project Gutenberg. Models such as Google PaLM have even used 2TB of high-quality book data to improve the model's knowledge reserves.

[0012] Social and Conversational Data: Conversational content from social media and forums is also an important supplement, such as Reddit and Ubuntu conversation data, which can enhance the model's conversational interaction capabilities. Some models from OpenAI, Deepseek, and Google also incorporate social media conversational data to make the model more closely resemble everyday communication contexts.

[0013] 2. Vertical professional data This type of data focuses on specific fields, is of high quality and highly professional, and is used to improve the accuracy of large models in specific scenarios. It usually needs to be obtained through authorization or screening.

[0014] Academic and research data: Academic papers are a core component, including datasets such as Cornell University's arXiv dataset, Sciencedirect in the publishing field, PubMed in the medical field, and CNKI and Wanfang in the Chinese field, covering multiple professional fields such as natural sciences, medicine, and engineering technology. Models like Zhipu Qingyan, which focus on academic capabilities, extensively incorporate this type of data to enhance their abilities in academic question answering and literature review. The large-scale model data proposed in this invention belongs to this type. The only difference is that it is applied to the application side.

[0015] Code data primarily comes from open-source platforms such as GitHub and GitLab, including BigCode's 6.4TB TheStack dataset and the StarCoder dataset, covering multiple mainstream programming languages. Models skilled in code generation, such as Deepseek, filter code from highly-rated open-source projects and remove low-quality duplicate content to improve performance on code-related tasks.

[0016] Publicly available data from institutions such as the National Digital Archives and the National Bureau of Statistics, including statistical data, policy documents, and industry reports, are also important sources of professional data. This data is highly authoritative and helps models understand policy regulations, industry trends, and other professional content. It is commonly used in training large-scale models in fields such as government affairs and finance.

[0017] 3. Enterprise-owned data This type of data is exclusive data accumulated by technology companies based on their own product ecosystem. It has uniqueness and scenario adaptability, and is also the key to the differentiated advantages of some large business models.

[0018] Data from its product ecosystems: ByteDance's Doubao will heavily utilize UGC and PGC content from its own platforms, such as short video scripts from Douyin, text and image news from Toutiao, and live streaming scripts from Xigua Video; Baidu Wenxin Yiyan will leverage its own product resources, such as structured knowledge from Baidu Encyclopedia and documents from Baidu Wenku, to enhance its knowledge service capabilities in Chinese-language scenarios; Tencent Yuanbao will also incorporate data from its ecosystem, such as content from WeChat official accounts and Tencent News materials.

[0019] Proprietary data in collaborative fields: Some companies collaborate with institutions in industries such as finance, healthcare, and law to obtain anonymized proprietary data. For example, Wenxin Yiyan incorporates professional corpora from collaborations with institutions in the financial and healthcare sectors, while Tencent Yuanbao uses anonymized enterprise customer data and game story text to adapt to scenarios such as financial services and game interactions.

[0020] The data above demonstrates that building large-scale models is never a simple matter, as the data sources are far from straightforward. This is especially true for laser additive manufacturing of high-temperature alloys, which involves complex parameters and is highly sensitive to cracks and defects. Furthermore, given the current scarcity of data, it's better to utilize artificial intelligence to achieve laser additive manufacturing of materials like high-temperature alloys. Computer simulation is a low-cost alternative for acquiring data; however, the accuracy of simulation models is affected by various simplifications such as material models, heat source models, and boundary conditions. Their predictions often deviate from physical reality. Directly using simulation data to train models may lead to a "garbage in, garbage out" problem, resulting in low reliability of the models in practical applications.

[0021] Therefore, how to efficiently and cost-effectively generate large-scale, high-confidence "process parameter-performance result" mapping data and build a reliable process knowledge database has become a key bottleneck in promoting the intelligentization of laser additive manufacturing and shortening the R&D cycle. Summary of the Invention

[0022] Purpose of the Invention: This invention aims to overcome the shortcomings of existing technologies and provide a method, system, and storage medium for establishing a database of laser additive manufacturing process parameters. This method, by constructing a closed-loop process of "simulation prediction - experimental verification - feedback correction," combines the advantages of low cost and high throughput of finite element simulation with the high reliability of physical experiments, enabling the systematic and efficient generation of high-quality structured data for training and verifying artificial intelligence models.

[0023] To achieve the above objectives, the present invention adopts the following technical solution: In a first aspect, the present invention provides a method for establishing a database of laser additive manufacturing process parameters, comprising the following steps: S1. Simulation data acquisition steps: For the target additive manufacturing material and process, establish the corresponding finite element simulation model, input the initial process parameter set for simulation calculation, and output simulation result data containing thermophysical field and mechanical field information; S2. Experimental data acquisition steps: Conduct actual laser additive manufacturing experiments based on the selected parameters in the initial process parameter set, and acquire the measured result data corresponding to the simulation result data through monitoring methods; S3. Data comparison and judgment steps: Compare the simulation result data with the measured result data, and make a consistency judgment based on the preset multiple-dimensional quantization error criteria. S4. Model correction and iteration steps: If the judgment result of step S3 is that all the quantization error criteria are not met, then one or more parameters in the finite element simulation model are adjusted, and step S1 is returned to perform iterative simulation and verification until the simulation and measured data meet all the quantization error criteria, and a corrected high-confidence simulation model is obtained. S5. Database generation step: Based on the corrected high-confidence simulation model, perform batch simulation calculations on the expanded process parameter space, associate and store the generated simulation data with the corresponding process parameters, and form a structured process parameter database.

[0024] Preferably, in step S1, establishing the finite element simulation model specifically includes: establishing the geometric model of the workpiece and the substrate and dividing the mesh; defining the thermophysical and mechanical parameters of the material; establishing the laser heat source model and defining its movement path; setting thermal boundary conditions and mechanical boundary conditions; and performing sequentially coupled thermal-stress transient analysis.

[0025] Preferably, the laser heat source model is a Gaussian surface heat source, a double ellipsoidal heat source, or a three-dimensional conical heat source; the material is a nickel-based high-temperature alloy, such as Inconel 718, Inconel 625, or DZ125 directional solidification alloy.

[0026] Preferably, in step S2, the monitoring methods include online monitoring and offline detection; the online monitoring includes using an infrared thermal imager or thermocouple to monitor the temperature field of the molten pool and using a high-speed camera to observe the morphology of the molten pool; the offline detection includes using X-ray diffraction to measure residual stress, using a three-dimensional optical scanner to measure deformation, and performing metallographic analysis to obtain the size and microstructure of the molten pool.

[0027] Preferably, in step S3, the quantization error criteria for the multiple dimensions include at least: (a) Criteria for thermal cycling curves: The relative error between the peak temperature of the monitoring points in the simulation and the actual measurement is ≤10%, and the cooling rate changes in the same trend; (b) Criteria for molten pool size: The relative error between the simulated high-temperature zone profile and the dimensional error of the melt width and / or melt depth obtained from metallographic analysis is ≤15%; (c) Residual stress criterion: The surface stress distribution cloud map obtained by simulation is consistent with the stress value measured by X-ray diffraction normal scan in terms of distribution trend, and the relative error of stress extrema is ≤20%; (d) Deformation criterion: The relative error between the maximum deformation of the vertical displacement field of the substrate calculated by simulation and the warping data obtained by three-dimensional scanning is ≤20%.

[0028] Preferably, in step S4, adjusting one or more parameters in the finite element simulation model includes: correcting the type or shape coefficient of the heat source model, adjusting the absorption rate of the material to the laser, supplementing or correcting the high-temperature constitutive relationship of the material, refining the convection or radiation boundary conditions, and adding the molten pool flow effect, etc.

[0029] Preferably, the laser additive manufacturing process is a laser powder bed melting process or a laser directional energy deposition process.

[0030] Secondly, the present invention provides a laser additive manufacturing process parameter database establishment system for implementing the above method, comprising: a simulation module, an experimental control and data acquisition module, a data comparison and correction decision module, and a database management module.

[0031] Thirdly, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-described method.

[0032] Beneficial effects: Compared with the prior art, the present invention has the following significant advantages: 1. High-quality and reliable data: Through closed-loop iteration of "experimental calibration simulation," the simulation model used to generate the database is ensured to have extremely high physical fidelity. The extended data generated based on this model has a much higher reliability than uncalibrated pure simulation data, providing a reliable data foundation for subsequent AI model training.

[0033] 2. High data generation efficiency and low cost: Compared to the "trial and error" method that relies entirely on physical experiments, this invention first uses simulation for preliminary screening and prediction, significantly reducing the number of expensive and time-consuming experiments. After obtaining the calibration model, computer simulation can be used to quickly and inexpensively explore a broad space of process parameters, generating massive amounts of structured data, thus solving the core problem of high cost in acquiring experimental data.

[0034] 3. The method is systematic and highly reproducible: It provides a complete standardized process from modeling, experimentation, comparison to calibration, and clarifies multi-dimensional and quantitative error judgment criteria, which makes the method reproducible and applicable to different materials, equipment and processes, and has good universality.

[0035] 4. Direct support for intelligent manufacturing: The established structured database can be directly used for training machine learning models (such as neural networks and random forests), enabling intelligent recommendation of process parameters, online prediction of forming quality, and automatic diagnosis of defects, significantly shortening the R&D cycle of new processes and materials (by more than 70%), reducing R&D costs, and improving the first-time success rate and stability of the manufacturing process.

[0036] 5. Significant knowledge accumulation value: Transforming tacit process experience and knowledge into explicit, structured digital assets, forming the enterprise's core process knowledge base, and preventing the loss of technical experience due to personnel turnover. Attached Figure Description Figure 1 This is a flowchart of the simulation prediction, experimental verification, and feedback optimization process for establishing a large-scale process data model as proposed in this invention. Figure 2 This is a comparison chart of the thermal cycling curves obtained from simulation and experimental testing in this invention; Figure 3 This is a comparison diagram of the molten pool size obtained from simulation and experimental testing in this invention; Figure 4 This is a comparison diagram of the deformation amount obtained from the simulated deterioration experiment in this invention. Detailed Implementation

[0037] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0038] The specific steps of this invention are as follows: It is mainly divided into three parts. The first part is the simulation part, which can use commercial analysis software such as Abaqus and Analysis. This invention will use Abaqus as an example. The core objective of this part is to obtain results such as thermal cycling curves, predicted temperature fields, thermal stress, deformation, and microstructure evolution.

[0039] The second part is the experimental verification section, the core objective of which is to obtain measured data and verify the reliability of the simulation.

[0040] The third part describes the results comparison and verification methods.

[0041] Specifically, in the first part of the operation, the first step is to establish the geometry (geometry of the sample or specimen) and the mesh. Specifically, this involves establishing a three-dimensional model of the substrate and the deposition layer, and using a fine mesh (refining the deposition area). Specifically, this involves using the "model change" method to simulate the addition of materials layer by layer, and the mesh needs to be refined to the scale of the light spot diameter.

[0042] Specifically, the first part of the operation involves establishing a material model, which specifically defines material parameters related to nickel-based alloys (such as Inconel 718, Inconel 625, etc.): thermophysical parameters (thermal conductivity, specific heat capacity, density); mechanical parameters (elastic modulus, yield strength, coefficient of thermal expansion); and optional items include adding a phase transformation model (such as precipitate kinetics). These data should be found in tool manuals or relevant publicly published scientific literature.

[0043] Specifically, in the first part of the operation, the heat source model is established, using a Gaussian surface heat source or a double ellipsoidal heat source; the heat source movement path is customized through the DFLUX or USDFLD subroutine. The heat source parameters (power, spot radius, absorptivity) must correspond to the experimental parameters; the path planning must match the actual scanning strategy (e.g., reciprocating scanning). Specifically, in the first part of the operation, boundary conditions are established, mainly including thermal boundaries: convection / radiation heat dissipation at the bottom of the substrate; and mechanical boundaries: constraining the degree of freedom at the bottom of the substrate. It is important to note that the high-temperature radiation heat dissipation coefficient needs to be set appropriately; and to avoid rigid displacement, weak springs or symmetrical constraints are used.

[0044] Specifically, in the first part of the operation, the solution setup adopts sequential coupled heat-stress analysis, mainly involving first performing transient thermal analysis (HeatTransfer); then importing the temperature field into static analysis (Static, General) to calculate stress, setting the incremental step and nonlinear solver. It is important to note that the large deformation effect (NLgeom) is enabled here, and the time increment is controlled to ensure thermal gradient convergence.

[0045] Specifically, in the first part of the operation, post-processing mainly involves extracting key results: temperature history curves (thermal cycles at monitoring points); residual stress distribution (principal stresses S11 and S22); deformation displacement field (U3 direction); geometry and mesh.

[0046] Specifically, in the experimental verification section of Part Two above, the core task is to obtain measured data and verify the reliability of the simulation. The following steps are required: Specifically, in the second part of the operation, the experiment is designed first, using the same process parameters as the simulation: laser power, scanning speed, spot diameter, layer thickness, scanning path; substrate pretreatment (polishing, cleaning, grinding, and whether preheating is required), etc. At this stage, the parameter combinations are optimized using orthogonal experimental design.

[0047] Specifically, in the second part of the operation, online monitoring is performed, using an infrared thermal imager or high-temperature thermocouples to record the temperature field of the molten pool. A high-speed camera observes the morphology of the molten pool. Time-temperature curves are recorded synchronously and compared with the simulated thermal cycle.

[0048] Specifically, in the second part of the operation, offline detection is performed, mainly including residual stress measurement: X-ray diffraction (XRD) or nanoindentation; deformation measurement: 3D optical scanner or laser displacement sensor; metallographic analysis: observation of the microstructure of the molten pool and heat-affected zone of the cross-sectional sample. It is important to note that the measurement location must correspond to the simulated monitoring point in space. Specifically, in the third part of the operation result comparison and verification method, a comparison of thermal cycling curves is required. The simulated monitoring point is the temperature-time curve of the monitoring point; the data monitoring point measured by Shuyan is thermocouple / thermal imager data. The peak temperature error of the monitoring point data of both is ≤10%, and the cooling rate trend is consistent, indicating that the data at that point is genuine.

[0049] Specifically, in the comparison and verification of the molten pool size in the third part of the operation results, it is necessary to monitor the contour of the high-temperature zone (e.g., the region ≥ liquidus temperature) in the simulation data; the experimental data is the melt width / deepness obtained from metallographic image analysis. A dimensional error of ≤15% between the two is considered true.

[0050] Specifically, in the comparison and verification of residual stress results in Part 3, the surface stress distribution cloud map in the simulation results is compared with the XRD line scan stress value obtained from the experiment. When the stress distribution trends of the two are consistent and the extreme value error is ≤20%, it is considered to meet the requirements.

[0051] Specifically, in the comparison and verification of deformation results in Part 3, the vertical displacement field of the substrate in the simulation results is compared with the actual measured three-dimensional scanning warping data. If the maximum deformation error between the two is ≤20%, it is considered to meet the requirements.

[0052] Example 1: Establishment of a database for the laser powder bed melting process of IN718 high-temperature alloy This embodiment uses the laser powder bed melting process of IN718 high-temperature alloy for aero-engines as an example to illustrate the specific application of this method.

[0053] S1: Simulation Data Acquisition 1. Geometry and Mesh: Using Abaqus software, a cubic single-pass multilayer model with dimensions of 10mm × 10mm × 5mm was created. The mesh in the deposition area was refined to 0.05mm, while the mesh in the substrate area was coarser.

[0054] 2. Material Model: Define the temperature-dependent properties of IN718 alloy: thermal conductivity, specific heat capacity, density, elastic modulus, Poisson's ratio, coefficient of thermal expansion, and yield strength. Data are sourced from the MMPDS manual and publicly available literature.

[0055] 3. Heat Source and Path: A double ellipsoidal heat source model is used, implemented via the DFLUX subroutine. The laser power is set to P = 300W, the scanning speed to V = 1000mm / s, and the spot diameter to d = 80μm. The scanning path is unidirectional, and the layer thickness is 50μm.

[0056] 4. Boundary conditions: A convective heat dissipation coefficient is applied to the bottom of the substrate, and radiative heat dissipation is applied to all outer surfaces. The bottom surface of the substrate is completely fixed and constrained.

[0057] 5. Solution and Post-processing: Perform transient thermo-mechanical sequential coupling analysis. Extract the temperature history curves (thermal cycles) of specific monitoring points, the residual stress field of the last layer (S11, S22), and the overall vertical displacement field (U3).

[0058] S2: Experimental Data Acquisition 1. Experimental design: IN718 cubic specimens were formed on a commercial LPBF equipment under the same process parameters as the simulation (300W, 1000mm / s, 80μm spot size, 50μm layer thickness).

[0059] 2. Online monitoring: A coaxially arranged high-temperature infrared thermal imager was used to record the temperature changes in the central region of the molten pool at a frequency of 10 kHz, and the experimental thermal cycling curve was obtained.

[0060] 3. Offline detection: Metallographic analysis: After wire cutting, mounting, polishing, and etching the sample, the width and depth of the molten pool are measured under a metallographic microscope.

[0061] Residual stress measurement: The residual stress on the upper surface of the sample along the scanning direction and perpendicular direction was measured using X-ray diffraction and the sin²ψ method.

[0062] Deformation measurement: The point cloud data of the upper surface of the formed sample is obtained using a 3D optical scanner and compared with the original CAD model to calculate the maximum warping deformation.

[0063] S3: Data Comparison and Criterion Determination Compare the data obtained from S1 and S2, and apply the following criteria (see...). Figures 2-4 (Illustration) Thermal cycling: The peak temperatures of the simulation and infrared thermometry are T_sim=2450°C and T_exp=2380°C, respectively, with a relative error of 2.9% (<10%). The heating and cooling curves are highly consistent, so the test is passed.

[0064] Melt pool size: The simulation predicted the melt width to be 115μm, while the metallographic measurement showed a melt width of 122μm. The relative error was 5.7% (<15%), and the result was deemed satisfactory.

[0065] Residual stress: The simulation predicted a maximum tensile stress of 520 MPa in the X direction on the surface, while the XRD measurement was 580 MPa. The relative error was 10.3% (<20%), and the stress distribution trend was consistent (compression at the edge and tension at the center), so the test was passed.

[0066] Deformation amount: The simulation predicted a maximum warpage of 0.28mm, while the 3D scan measured 0.35mm, with a relative error of 20% (equal to the threshold). After manual review, the trend was consistent, and it was barely deemed acceptable, but optimization in the next iteration is recommended.

[0067] S4: Model Correction and Iteration Since the deformation error was at a critical value, it was decided to fine-tune the model to improve accuracy. The elastic modulus of the material in the high-temperature range (>1200°C) was reduced by 5% based on the latest literature data. Steps S1-S3 were then repeated using the corrected model.

[0068] After the second iteration, the deformation prediction error decreased to 15%, and all criteria were better met. At this point, the simulation model was considered to have been effectively corrected and could be used as a high-confidence model.

[0069] S5: Database Generation Using a calibrated high-confidence simulation model, simulations with different combinations of process parameters are automatically run in batches. For example, full-factor or Latin hypercube sampling simulations are performed within the parameter space: power P=[200,250,300,350,400]W, speed V=[600,800,1000,1200,1400]mm / s. Each simulation outputs corresponding data such as thermal cycle, melt pool size, residual stress, and deformation. These input-output pairs are structured and stored in an SQL or NoSQL database, forming an initial version of the IN718 alloy LPBF process database. This database can be directly used to train machine learning models for predicting forming quality or recommending processes.

[0070] In summary, the simulation-verification process data of Case 1 - IN718 high-temperature alloy laser selective melting is presented in a large-scale model. A combustion chamber component of a certain aero-engine is manufactured using IN718 high-temperature alloy. Traditional forging processes result in low material utilization (<30%) and long development cycles. Laser powder bed fusion (LPBF) technology can significantly improve material utilization (>95%) and achieve integrated manufacturing of complex structures. However, the LPBF process is prone to: Porosity, lack of fusion defects High residual stress leads to deformation and cracking. Anisotropic microstructures By using a closed loop of "simulation prediction - experimental verification - feedback optimization", the optimal process parameter window can be quickly determined, achieving the following: Density > 99.5% Residual stress <500MPa Mechanical properties meet AMS5662 standard First, the simulation model is built by setting the following parameters in Abaqus: Laser power: [200,250,300,350,400],#W "Scanning speed":[600,800,1000,1200],#mm / s "Scan spacing": 0.1, #mm Layer thickness: 0.05 mm "Spot diameter": 0.08, #mm Scanning strategy: "Stripes rotated 67°" Key results predicted:

[0071] Simulation-Experiment Comparison Analysis

[0072] Model correction measures 1. Heat source model correction: Change the Gaussian heat source to a double ellipsoidal heat source, and set the rear length coefficient to 1.5; 2. Absorption rate adjustment: corrected from 0.35 to 0.33 (considering the plasma effect of the molten pool); 3. Improved material parameters: Supplemented creep data for IN718 at 1300-1400°C; 4. Refine boundary conditions: Increase the Marangoni convective heat transfer coefficient of the molten pool; Improved simulation accuracy after correction Temperature field prediction error decreased from an average of 12% to 5%; The error in residual stress prediction decreased from 15% to 8%; The error in predicting the molten pool morphology decreased from 8% to 4%; Based on the simulation results, the following results can be obtained. Based on the analysis of the temperature field and stress field: 1. Preheating the substrate to 200°C can reduce the thermal gradient by approximately 40%; 2. The partitioned scanning strategy can reduce peak stress by approximately 30%; 3. Contour scanning parameters should be 15-20% lower in power than fill scanning.

[0073] Experimental verification, Based on simulation prediction, 8 sets of experiments were designed: The main results of the actual measurement are as follows: Key measured data (Group A2: 300W / 1000mm / s / 200°C preheating) Actual measurement results = { Density: 99.3%, measured by Archimedes method Surface roughness: Ra 12.5, #μm "Molten pool size":{ Width: 0.122 mm, metallographic measurement Depth: 0.049, #mm }, "Residual stress":{ "Surface X direction": 580, #MPa, XRD measurement "Surface Y direction": 610, #MPa Internal average: 520 MPa, layer-by-layer peeling method }, "Heat distortion":{ Maximum warpage: 0.35 mm, 3D scan Average shrinkage: 0.12,#% } } Iterative optimization and final verification Re-optimization based on the modified model The corrected simulation prediction optimal parameters are: Power: 310W Scanning speed: 1050mm / s Scanning interval: 0.09mm Layer thickness: 0.04mm Substrate preheating: 250°C Scanning strategy: Partition scanning + contour compensation Final experimental verification results

[0074] Microstructure analysis Grain morphology: Predominantly epitaxial columnar crystals, with an average width of 45 μm. Precipitated phases: uniformly distributed γ' and γ'' phases, with a size of 20-50 nm. Molten pool boundary: No obvious elemental segregation, no continuous Laves phase network structure. Summary of Closed-Loop Optimization Results Efficiency Improvement

[0075] Cost Savings Analysis Direct cost savings: approximately 120,000 yuan (materials, machine time, testing). Indirect cost savings: shorten product time to market by 3-4 months Quality cost reduction: Scrap rate decreased from 25% to 8%. Based on the establishment of this large model, the following significant technological advancements have been achieved: Through a closed-loop process of "simulation prediction - experimental verification - feedback optimization", the rapid development and optimization of the LPBF process for IN718 high-temperature alloy was achieved. 1. Technological Achievements: Obtaining a process window that combines high density, low residual stress, and excellent mechanical properties. 2. Methodological Value: This study validated the high efficiency of simulation-driven R&D, reducing the development cycle by more than 70%. 3. Scalability potential: This closed-loop method can be extended to the development of additive manufacturing processes for other difficult-to-process materials.

[0076] Example 2: Establishment of a database for laser-directed energy deposition repair process of CoCrW coating on DZ125 blades This embodiment demonstrates the application of the method of the present invention in the complex scenario of laser-directed energy deposition repair of dissimilar materials. The process is similar to that of Embodiment 1, but the specificity is highlighted.

[0077] S1: A multi-material finite element model was established, including a DZ125 substrate and a CoCrW coating, considering the differences in their thermal properties. A 3D conical heat source was used to simulate the laser energy distribution under powder feeding conditions.

[0078] S2: L-DED repair experiments were conducted on actual blade test blocks, with a focus on monitoring the interfacial bonding zone. The temperature and morphology of the molten pool were monitored online, while interfacial cracks, dilution rate, interfacial bonding strength (tensile test), and the width of the heat-affected zone were detected offline.

[0079] S3: In addition to the conventional criteria, add interface performance criteria: the maximum principal stress of the interface predicted by simulation is correlated with the crack initiation tendency observed in experiments; the error between the predicted composition of the melt pool overlap zone (used to calculate the dilution rate) and the composition measured by electron probe is ≤15%.

[0080] S4: The first round of simulations revealed that the predicted interfacial stress (850 MPa) was significantly higher than the experimental safety threshold (500 MPa), and the microcracks that appeared in the experiment were not predicted. Corrective measures included: adding the high-temperature creep constitutive relation of DZ125 to the material model to simulate stress relaxation; and correcting the surface tension temperature coefficient of the CoCrW molten pool to more accurately simulate Marangoni convection. After two iterations, the predicted interfacial stress was reduced to 400 MPa, and the crack risk region could be qualitatively predicted, consistent with the experimental results.

[0081] S5: Based on the calibration model, extended simulations are performed on L-DED-specific parameters such as powder feeding rate, scanning path, and preheating temperature to establish an L-DED repair process database for the "DZ125-CoCrW" material system, which is used to optimize the repair process and ensure repair quality.

[0082] In summary, the establishment of large-scale model data for laser-directed energy deposition repair of CoCrW wear-resistant coating at the tip of DZ125 blades in Example 2 was successful. After long-term high-temperature service, the DZ125 directionally solidified superalloy blades of aero-engines experience severe wear (0.5-2 mm) in the blade tip area, affecting aerodynamic performance. Traditional repair methods (such as TIG welding) have problems such as large heat-affected zones, grain coarsening, and high residual stress.

[0083] technical route Laser-directed energy deposition (L-DED) technology is used to deposit CoCrW wear-resistant alloy (Stellite6 or 21 series) on worn blade tips for splicing repair, which must meet the following requirements: Metallurgical quality: No cracks or pores at the interface. Organizational control: Avoid excessive remelting of columnar crystals in the DZ125 matrix. Stress control: Residual stress <500MPa to prevent blade deformation. Service performance: Wear resistance is improved by more than 3 times, and it has good resistance to high-temperature oxidation. Phase 1: Multiphysics Simulation and Prediction Simulation model establishment Python Material parameter settings Material properties={ "DZ125 substrate":{ Liquidus temperature: 1360°C #°C Solidus temperature: 1260°C #°C Thermal conductivity: "Temperature-related function", Coefficient of thermal expansion: ≈16.5×10⁻ 6 / °C, / °C (20-1000 °C) "Elastic modulus": "Temperature-dependent function" }, "CoCrW coating":{ Liquidus temperature: 1410°C #°C Solidus temperature: 1290°C #°C Thermal conductivity: 14.8 W / m K Coefficient of thermal expansion: 14.2e-6 # / °C Elastic modulus: 230 GPa } } Thermal-mechanical-flow multi-field coupling model Model configuration = { "Geometric Model": "Simplified CAD model of the actual blade, tip region", "Mesh Generation": "Adaptive encryption, 0.1mm mesh at the interface", Analysis type: ["Transient thermal analysis", "Fluid-structure interaction", "Phase change analysis"] Boundary conditions:{ "Thermal boundary": "Convection + radiation, considering blade cooling channels", "Mechanical Boundary": "Blade Tenon Fixing Constraint" "Flow field": "Molten pool Marangoni convection" } } Sensitivity analysis of key process parameters

[0084] Simulation prediction results and optimization suggestions The first round of simulations revealed key issues: Crack sensitivity: The difference in thermal expansion coefficients between DZ125 and CoCrW leads to interfacial stress concentration, with the maximum principal stress reaching 850 MPa. Dilution rate control: The ideal dilution rate should be controlled between 10-20%. Too low a rate results in poor bonding, while too high a rate damages the matrix properties. Heat-affected zone: Overheated matrix zone > Simulation optimization suggestions: Phase Two: Experimental Verification and Characterization Experimental Design Based on simulation recommendations, three key experimental groups were designed:

[0085] Experimental Results and Key Data Python Experimental results = { "Group G1 (No Transition Layer)":{ "Macroscopic Defects": "Two longitudinal cracks at the interface, 3-5mm in length". "Dilution rate": 28.5 %, metallographic measurement "Interfacial bonding strength": 320, MPa, tensile test Residual stress: 680, MPa, XRD measurement "Width of heat-affected zone": 2.3 mm }, "G2 Group (Gradual Transition)":{ "Macroscopic Defects": "No cracks, a small amount of porosity (<1%)", Dilution rate: 18.2 #% "Interfacial bonding strength": 450, MPa Residual stress: 520, MPa "Width of heat-affected zone": 1.8 mm }, "G3 Group (Optimized Parameters)":{ "Macroscopic Defects": "No cracks, no pores" "Dilution rate": 15.6, #% "Interfacial bonding strength": 510, MPa Residual stress: 380, MPa "Width of heat-affected zone": 1.2 mm }} Microstructure analysis results

[0086] Phase 3: Simulation - Experimental Comparison and Model Correction Key Indicator Comparison Analysis

[0087] Model correction strategy Add a high-temperature creep constitutive model for DZ125 (Norton-Bailey equations). Corrected surface tension temperature coefficient of CoCrW molten pool (-0.35×10⁻³N / m·K) Change from Gaussian heat source to 3D cone heat source Considering the shielding effect of the powder flow on the laser (efficiency coefficient 0.85). Add actual internal cooling channels for convective heat transfer in the blades Consider forced convection of protective gas (argon, flow rate 15 L / min). Corrected model validation The corrected model's prediction accuracy for the second set of validation experiments:

[0088] Phase 4: Final Optimization and Actual Blade Repair Final optimization based on the modified model Optimal combination of process parameters: Laser power: 960±20W Scanning speed: 11.5 mm / s Powder delivery rate: 11.2g / min Spot diameter: 2.8mm Preheating temperature: 450°C Interlayer temperature control: Start the next layer when the temperature is below 200°C. Scanning path: spiral outwards, overlap rate 35% Gradient transition layer design: Text

[0089] Actual blade repair results Repairing 5 DZ125 high-pressure turbine blades (actual service parts):

[0090] Performance test results

[0091] VI. Cost-Benefit Analysis and Techno-Economic Efficiency Repair cost comparison

[0092] Life cycle benefits Direct economic benefits: 3,100 yuan is saved per piece, and based on 200 pieces repaired per year, the annual savings amount to 620,000 yuan; Indirect benefits: Reduced spare parts procurement and extended service life; Technical benefits: Establishing a digital repair process package reduces the repair development cycle for similar parts by 70%. Successfully achieved the following through a complete closed loop of "simulation prediction - experimental verification - feedback optimization": Figure 1 The system architecture for implementing the above method is demonstrated. This system integrates simulation software, device controllers, data acquisition equipment, a central database, and management software, enabling semi-automatic or fully automatic closed-loop data production processes.

[0093] Obviously, the above embodiments are merely illustrative examples for clear explanation and are not intended to limit the implementation. Those skilled in the art will recognize that other variations or modifications can be made based on the above description. It is neither necessary nor possible to exhaustively list all possible implementations here. However, obvious variations or modifications derived therefrom are still within the scope of protection of this invention.

Claims

1. A method for establishing a database of laser additive manufacturing process parameters, characterized in that, Includes the following steps: S1. Simulation data acquisition steps: For the target additive manufacturing material and process, establish the corresponding finite element simulation model, input the initial process parameter set for simulation calculation, and output simulation result data containing thermophysical field and mechanical field information; S2. Experimental data acquisition steps: Based on the selected parameters in the initial process parameter set, conduct actual laser additive manufacturing experiments, and obtain the measured result data corresponding to the simulation result data through monitoring methods; S3. Data comparison and judgment steps: Compare the simulation result data with the measured result data, and make a consistency judgment based on the preset multiple-dimensional quantization error criteria. S4. Model correction and iteration steps: If the judgment result of step S3 is that all the quantization error criteria are not met, then one or more parameters in the finite element simulation model are adjusted, and step S1 is returned to perform iterative simulation and verification until the simulation and measured data meet all the quantization error criteria, and a corrected high-confidence simulation model is obtained. S5. Database generation step: Based on the corrected high-confidence simulation model, perform batch simulation calculations on the expanded process parameter space, associate and store the generated simulation data with the corresponding process parameters, and form a structured process parameter database.

2. The method for establishing a laser additive manufacturing process parameter database according to claim 1, characterized in that, In step S1, establishing the finite element simulation model specifically includes: establishing the geometric model of the workpiece and the substrate and dividing it into meshes; defining the thermophysical and mechanical parameters of the material; establishing the laser heat source model and defining its movement path; setting thermal boundary conditions and mechanical boundary conditions; and performing sequentially coupled thermal-stress transient analysis.

3. The method for establishing a laser additive manufacturing process parameter database according to claim 2, characterized in that, The laser heat source model is a Gaussian surface heat source, a double ellipsoidal heat source, or a three-dimensional conical heat source; the material is a nickel-based high-temperature alloy.

4. The method for establishing a laser additive manufacturing process parameter database according to claim 1, characterized in that, In step S2, the monitoring methods include online monitoring and offline detection; the online monitoring includes using an infrared thermal imager or thermocouple to monitor the temperature field of the molten pool and using a high-speed camera to observe the morphology of the molten pool; the offline detection includes using X-ray diffraction to measure residual stress, using a three-dimensional optical scanner to measure deformation, and performing metallographic analysis to obtain the size and microstructure of the molten pool.

5. The method for establishing a laser additive manufacturing process parameter database according to claim 1, characterized in that, In step S3, the quantization error criteria of the multiple dimensions include at least: (a) Criteria for thermal cycling curves: The relative error between the peak temperature of the monitoring points in the simulation and the actual measurement is ≤10%, and the cooling rate changes in the same trend; (b) Criteria for molten pool size: The relative error between the simulated high-temperature zone profile and the dimensional error of the melt width and / or melt depth obtained from metallographic analysis is ≤15%; (c) Residual stress criterion: The surface stress distribution cloud map obtained by simulation is consistent with the stress value measured by X-ray diffraction normal scan in terms of distribution trend, and the relative error of stress extrema is ≤20%; (d) Deformation criterion: The relative error between the maximum deformation of the vertical displacement field of the substrate calculated by simulation and the warping data obtained by three-dimensional scanning is ≤20%.

6. The method for establishing a laser additive manufacturing process parameter database according to claim 5, characterized in that, In step S4, adjusting one or more parameters in the finite element simulation model includes: correcting the type or shape factor of the heat source model, adjusting the absorption rate of the material to the laser, supplementing or correcting the high-temperature constitutive relationship of the material, and refining the convection or radiation boundary conditions.

7. The method for establishing a laser additive manufacturing process parameter database according to claim 1, characterized in that, The laser additive manufacturing process is either laser powder bed melting or laser directional energy deposition.

8. The method for establishing a laser additive manufacturing process parameter database according to claim 7, characterized in that, In step S5, the data entries in the structured process parameter database include: laser power, scanning speed, scanning spacing, layer thickness, spot diameter, scanning strategy, preheating temperature process parameters, as well as associated density, melt pool size, residual stress, deformation amount, and microstructure characteristics result parameters.

9. A laser additive manufacturing process parameter database establishment system for implementing the method for establishing a laser additive manufacturing process parameter database according to any one of claims 1-8, characterized in that, include: The simulation module is used to build and run finite element simulation models and output simulation results data; The experimental control and data acquisition module is used to control the laser additive manufacturing equipment to operate according to specified process parameters and integrates various sensors to collect measured data. The data comparison and correction decision module is used to receive simulation and measured data, automatically compare and judge them according to preset quantization error criteria, and generate simulation model parameter correction suggestions based on the judgment results. The database management module is used to store, manage, and associate the iteratively corrected set of process parameters and their corresponding performance results data to form a queryable and scalable structured database.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the program implements the steps of the method for establishing a laser additive manufacturing process parameter database as described in any one of claims 1-8.

Images