Precise additive manufacturing micro-channel heat exchanger forming quality evaluation method
By extracting representative cell models and conducting multi-dimensional tests, combined with machine learning models, the problem of traditional methods being unable to quickly evaluate the quality of microchannel heat exchangers was solved, achieving low-cost and rapid quality evaluation and process optimization.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-09
- Publication Date
- 2026-04-10
AI Technical Summary
Traditional methods are difficult to evaluate the internal forming quality of microchannel heat exchangers in a non-destructive, low-cost, and rapid manner, especially the geometric accuracy and surface roughness of complex thin-walled structures. Furthermore, traditional testing methods cannot predict the applicability of LPBF process in the early stages.
By extracting representative cell models and performing precision additive manufacturing, combined with laser confocal microscopy, industrial CT scanning, metallographic sample preparation, and mechanical property testing, a machine learning model is established to achieve non-destructive quality evaluation of microchannel heat exchangers.
It enables low-cost and rapid quality evaluation of microchannel heat exchangers, which can represent the internal forming quality of the overall structure, guide process optimization and predict performance, and reduce R&D cycle and cost.
Smart Images
Figure CN121835355A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of additive manufacturing quality detection, in particular to a precision additive manufacturing micro-channel heat exchanger forming quality evaluation method. BACKGROUND
[0002] Micro-channel heat exchangers are generally defined as heat exchangers with a channel equivalent diameter of 10-1000 μm. Due to its high specific surface area, compact structure and high efficiency heat exchange performance, it has been widely used in high-end fields such as aerospace, new energy vehicles, precision electronic equipment, etc. To improve the heat exchange efficiency, complex topological structures such as three-period minimal surface (TPMS) are generally used inside this type of heat exchanger, and the characteristic size is usually in the range of hundreds of microns to millimeters. The wall thickness is thin, the porosity is high, and the geometry is complex, so it is difficult to achieve integrated manufacturing by traditional subtractive or casting forming methods.
[0003] In recent years, metal additive manufacturing technologies such as laser powder bed fusion (LPBF) have become an ideal process route for manufacturing complex micro-channel heat exchangers due to their high forming freedom and near-net shaping capability. However, this technology still faces major challenges when preparing ultra-thin wall features (usually with a thickness of less than 200-300 μm). For example, it is very difficult for LPBF process to manufacture thin-walled structures with high quality and no defects. Thin-walled structures are prone to warping and cracking. On the other hand, there are many internal curves in the micro-channel, and it is difficult to completely clean the powder after printing. The residual powder attached to the inner wall will significantly increase the inner surface roughness, thereby reducing the heat exchange performance. These problems have driven the traditional metal additive manufacturing technology to develop towards a more precise additive manufacturing technology.
[0004] At the same time, the development of micro-channel heat exchanger precision additive manufacturing technology has also given rise to new problems. For micro-channel heat exchangers with a large number of internal channels and ultra-thin wall structures, the geometric size tolerance and surface roughness of the channels have a significant impact on the heat exchange efficiency. However, traditional probe or visual detection methods cannot directly measure the surface state and wall thickness of closed internal micro-channels, making it difficult to obtain key indicators such as the inner surface quality of the heat exchanger, the geometric precision of the thin wall, the material density and the mechanical properties, and unable to form a closed loop guidance and optimization of the forming process. In addition, the industry's system evaluation method for heat exchanger forming quality is still not perfect. The traditional verification method requires the whole piece to be printed, and then the internal details of the large-size part can be observed by industrial CT (which is difficult to observe) or section detection (the sample is destroyed and scrapped), which has a long cycle and high cost, and cannot early predict the applicability of the LPBF process. The TPMS structure inside the heat exchanger is diverse, and there is a lack of unified and repeatable sampling and characterization methods, which makes it difficult to compare the quality of different product batches and different equipment.
[0005] Therefore, it is urgent to propose a non-integral destructive, low-cost and short-cycle additive manufacturing heat exchanger forming quality standardization evaluation method, which can represent the overall structure characteristics, so as to comprehensively evaluate the inner and outer surface quality, size accuracy, density and mechanical properties of the heat exchanger product, and realize the unity of process verification before printing, quality monitoring during printing and performance prediction after printing. SUMMARY
[0006] The application provides a precision additive manufacturing micro-channel heat exchanger forming quality evaluation method, which completes representative evaluation of the internal structure forming quality of the actual heat exchanger component without damaging the actual heat exchanger component, and solves the problem that internal forming defects of complex thin-walled structural parts are difficult to quantify.
[0007] To achieve the above object, the technical scheme of the application is as follows:
[0008] S1 obtains an original three-dimensional STL model of a heat exchanger to be manufactured, and the model inside contains a three-period minimal surface (TPMS) cell structure arranged periodically;
[0009] S2 cuts a representative cycle unit cell from the original STL model, adds a simple thin-walled lug with the same characteristic wall thickness to the side surface of the unit cell, and merges them into one part to derive a local test part model, and the representative unit cell is consistent with the overall structure of the heat exchanger in terms of topological type, wall thickness distribution, porosity and curvature characteristics;
[0010] S3 uses a laser powder bed fusion (LPBF) precision additive manufacturing equipment to prepare the local test part model, obtains a metallographic cell sample, and performs wire cutting separation on the sample;
[0011] S4 performs multi-dimensional performance testing on the metallographic cell sample, including:
[0012] a. Surface roughness test: a laser confocal microscope or a white light interferometer is used to non-contact scan the original state cell surface to obtain roughness parameters such as Ra, Rz and Sa, which are used to evaluate the forming surface quality;
[0013] b. Industrial CT scanning: high-resolution X-ray three-dimensional imaging is performed on the original state cell sample to extract actual wall thickness distribution, geometric deviation and internal defect information, which are used to evaluate the thin-walled forming accuracy;
[0014] c. Metallographic sample preparation and observation: after embedding, polishing and polishing the cell sample, an optical microscope or a scanning electron microscope is used to observe the cross section, measure the minimum wall thickness, average wall thickness and local density, and identify defects such as incomplete fusion, pores and cracks;
[0015] d.Mechanical property test: Quasi-static compression test is performed on the unit cell sample to obtain the elastic modulus, yield strength, peak stress and energy absorption capacity for evaluating the structural bearing performance.
[0016] S5 Based on the above test results, a "process parameter-unit cell quality-heat exchanger performance" correlation model is established by machine learning method to guide the LPBF process optimization and finished product quality prediction of actual heat exchanger components.
[0017] Preferably, the intercept position of the representative unit cell is located in the full thin-walled area inside the heat exchanger, the contour size is 1 unit cell period, and the surface cutting position is located in the symmetric neutral surface of the unit cell structure to avoid boundary effect interference with the test results.
[0018] Preferably, the measurement area of the surface roughness test should be selected at the smooth part of the metallographic TPMS unit cell surface for subsequent leveling. The roughness parameter indicators measured under the 50x lens of the laser confocal microscope test objective should meet: Ra≤5μm, Rz≤10μm, Sa≤10μm.
[0019] Preferably, the resolution of the industrial CT scan is ≤5μm, a local magnification scanning mode is adopted, and geometric registration is performed based on the STL model. The minimum wall thickness positive and negative deviation ≤20μm is quantitatively output through the deviation chromatogram.
[0020] Preferably, the metallographic unit cell sample is cold-mounted or hot-mounted, and after step-by-step grinding (using fine sandpaper), 3μm and 1μm particle size diamond polishing paste is used for polishing in turn, and finally 0.25μm particle size polishing liquid is used for final polishing to obtain a mirror surface level metallographic sample; ImageJ image processing software is used for image binarization processing of the unit cell sample observation, and the density = (solid pixel number / total pixel number) × 100% is calculated. There should be no obvious unfused area and macroscopic crack in the solid area, and the density ≥99.5%.
[0021] Preferably, the compression test adopts displacement control mode, the loading rate is 0.5-1mm / min, the sample height-diameter ratio is 1:1~1.5:1, and the upper and lower end faces are ground to ensure that the parallelism is ≤0.01mm. The stress-strain curve is used to extract the platform stress and densification strain.
[0022] Preferably, the "process parameter-unit cell quality-heat exchanger performance" correlation model is constructed by machine learning algorithm, the input variables include laser power, scanning speed, layer thickness, and interpass distance, and the output variables include roughness Ra, Rz, Sa, minimum wall thickness t min , density ρ rel , compression strength σ yAfter the model is trained, it is used to predict the overall heat exchanger performance under different processes, enabling rapid selection of process windows.
[0023] Preferably, the TPMS structure is of the Gyroid, Diamond, or I-WPSchwarz-P type, and the cell size is 5~20mm, with a cell feature wall thickness of 0.1~0.8mm.
[0024] Preferably, the raw materials for the laser powder bed melting (LPBF) manufacturing of the microchannel heat exchanger are selected from one of titanium alloys, aluminum alloys, copper alloys, stainless steel or nickel-based high-temperature alloys, and the material is a spherical pre-alloyed powder with excellent spreadability and density during forming.
[0025] The beneficial effects of this invention are as follows:
[0026] 1. Significantly reduces the quality evaluation cost and R&D cycle of microchannel heat exchangers. Evaluation can be completed by printing only a single micro-sample block of a single cell, saving the lengthy printing and post-processing costs of the entire heat exchanger.
[0027] 2. The single-cell structure can represent the typical heat exchange area of a microchannel heat exchanger. Through four-dimensional data (roughness, wall thickness deviation, density, and mechanical properties), non-destructive prediction and evaluation of internal structural quality can be achieved by "representing the large with the small".
[0028] 3. Based on the same cell sample, it can complete the detection requirements of four aspects: internal and external surface quality, geometric accuracy, density and mechanical properties at one time, avoiding the process window drift caused by the traditional "multiple batches and multiple samples", and providing complete input for machine learning models.
[0029] 4. First, use preset indicators to quickly determine whether the process is qualified or needs to be adjusted. Then, use machine learning to establish a "process-quality-performance" correlation model with R²≥0.85. This model can be directly extended to heat exchangers of any size with the same material and configuration, achieving second-level prediction of process windows.
[0030] 5. The method of this invention is not limited to specific materials or specific heat exchanger structures. It can be extended to additive manufacturing heat exchanger products of various material systems such as titanium alloy, high-temperature alloy, and aluminum alloy, as well as various TPMS topology-optimized structure heat exchangers such as Gyroid, Diamond, and I-WP. All steps can establish a standardized quality evaluation process for additive manufacturing heat exchangers, providing a unified quality evaluation basis for the industry.
[0031] 6. By rapidly iterating the laser powder bed melting process using "small cells", the cost of adjusting parameters such as laser power, scanning speed, and layer thickness is reduced, which promotes the development of microchannel heat exchangers towards higher specific surface area and thinner wall thickness (0.1mm level), significantly improving heat exchange efficiency. Attached Figure Description
[0032] Figure 1 A schematic diagram of extracting a single cell from the STL model of the overall heat exchanger;
[0033] Figure 2 This is a schematic diagram of the implementation process of the present invention;
[0034] Figure 3 A schematic diagram of a printed monomer metallographic cell sample;
[0035] Figure 4 Grayscale image of the TPMS cell surface scanning region extracted by laser confocal microscopy;
[0036] Figure 5 Schematic diagrams of hot-mounted (left) and cold-mounted (left) metallographic cell samples;
[0037] Figure 6 This is a metallographic diagram of a cell-unit sample under an optical microscope.
[0038] Figure 7 To obtain the 3D image of the cell sample from an industrial CT scan;
[0039] Figure 8 A schematic diagram of the cross-sectional morphology in a 3D industrial CT scan of a cell sample;
[0040] Figure 9 This is a schematic diagram of the cell sample compression test process;
[0041] Figure 10 The diagram shows the compressive strain and stress changes of the cell sample under a loading rate of 0.5 min / min during a compression experiment at room temperature. Detailed Implementation
[0042] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.
[0043] 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.
[0044] Example 1:
[0045] S1 Model Preparation
[0046] Import the target GH3536 heat exchanger integral STL into Materialise Magics software. Its interior is a 5mm periodic Gyroid cell array with a designed wall thickness of 0.20mm and a porosity of 75%.
[0047] Select the middle region and cut out a 1×1×1 cell cube along each of the
[100] ,
[010] , and
[001] periods; place the cutting surface on the Gyroid neutral plane to avoid boundary effects.
[0048] Thin-walled lugs with a thickness of 0.20 mm and a height of 2 mm are added to opposite outer surfaces of the six faces of the cell for subsequent clamping and alignment. After merging, the lugs are exported as "Gyroid-GH3536-UnitCell.stl".
[0049] S2 Powder and Equipment
[0050] GH3536 pre-alloyed powder was atomized by vacuum induction, with a particle size of 0-25μm, D50=12μm, and a loose packing density of 4.85g / cm³ (≈52% of theoretical density).
[0051] Vacuum dry at 100℃ for 6 hours before printing, with an oxygen content ≤200ppm.
[0052] LPBF device: LiM-X260A, equipped with a 400W single-mode fiber laser with a focal spot diameter of 50μm; the substrate is 316L stainless steel, and the preheating temperature is 80℃.
[0053] S3 Input Process Parameters LPBF Forming
[0054] Laser power 85W, scanning speed 800mm / s, layer thickness 20μm, track spacing 60μm, strip width 2mm, interlayer rotation angle 67°; argon protection throughout the process, oxygen in the forming chamber ≤800ppm.
[0055] S4 Forming and Sampling
[0056] The printed volume is 8mm×8mm×8mm. The sample was removed by wire EDM, and there were no macroscopic cracks on the surface.
[0057] S5 Multidimensional Test
[0058] a) Surface roughness: The surface roughness was measured in the Gyroid saddle point flat area using a laser confocal microscope (50× objective lens) with Ra=3.2μm, Rz=7.8μm, and Sa=6.5μm, all of which meet the requirements of Ra≤5μm, Rz≤10μm, and Sa≤10μm in claim 3 of the abstract.
[0059] b) Industrial CT scan: After cell-level μCT scan, registration is performed using STL as the reference. The minimum wall thickness is 0.192 mm, and the absolute deviation from the design value of 0.20 mm is +4 / -6 μm. Positive and negative deviations ≤20 μm are acceptable.
[0060] c) Metallographic density: cold mounting → progressive grinding to 2000# → 3μm and 1μm diamond polishing → 0.25μm silicone suspension final polishing; ImageJ binarization of 10 500× images, average density 99.93%, no >5μm unfused or cracked areas observed.
[0061] d) Compression performance: Sample height 7.5 mm, diameter 5 mm (height-to-diameter ratio 1.5:1), parallelism of ground upper and lower end faces 0.008 mm; displacement control of electronic universal testing machine 0.5 mm / min. Measured at room temperature:
[0062] Elastic modulus E = 205 GPa, yield strength σ y =820MPa, peak stress 980MPa, densification strain ε dens =0.42, with a deviation of <3% from the sampling data of the same process integral heat exchanger.
[0063] S6 Machine Learning Modeling
[0064] Forty-five sets of small cell data with different powers (55-105W), speeds (600-1000mms⁻¹), and layer thicknesses (15-25μm) were used as input, with Ra and t as the data. min ρ rel σ y For the output, random forest regression was used, and 10-fold cross-validation yielded R²=0.87, which meets the requirement of R²≥0.85 in claim 7 of the abstract. The model predicted the optimal window: power 85W, speed 850mm / s⁻¹, layer thickness 20μm, with an error of <2% compared to the actual measurements in this embodiment.
[0065] S7 Performance Verification
[0066] The entire 200mm×200mm×210mm Gyroid-GH3536 heat exchanger was reprinted according to the optimized parameters given by the model. Industrial CT sampling inspection showed an average wall thickness deviation of 8μm, a density of 99.91%, and no air leakage after holding at 0.6MPa pressure for 20 minutes. This proves that the "representative cell sampling-evaluation-prediction" process proposed in this embodiment can efficiently guide the precision additive manufacturing and quality assurance of complex high-temperature alloy microchannel heat exchangers.
[0067] Example 2:
[0068] S1 Model Preparation
[0069] Import the target TA15 heat exchanger into Materialise Magics as a whole STL, with an internal 10mm periodic Diamond (Schwarz-D) cell array, a wall thickness of 0.35mm, and a porosity of 70%.
[0070] Select the central region of the heat exchanger and cut a 1×1×1 cell cube along each of the
[100] ,
[010] , and
[001] cycles; place the cut surface on the Diamond neutral plane to avoid boundary effects.
[0071] Thin-walled lugs, 0.35 mm thick and 3 mm high, are added to the outer surfaces of the six sides of the cell for subsequent clamping and alignment. After merging, the file is exported as "Diamond-Ti65-UnitCell.stl".
[0072] S2 Powder and Equipment
[0073] The Ti65 (Ti-6Al-6V-2Sn-2Zr-2Mo-1Cr-1Fe) pre-alloyed powder was atomized by electrode induction. The particle size was 0-30μm, D50=15μm, and the loose packing density was 2.75g / cm³ (≈58% of the theoretical density).
[0074] Vacuum dry at 120℃ for 8 hours before printing, with an oxygen content ≤150ppm.
[0075] Equipment: LiM-X250A, equipped with a 500W single-mode fiber laser with a focal spot diameter of 55μm; the substrate is also Ti65, and the preheating temperature is 200℃.
[0076] S3 Input Process Parameters LPBF Forming
[0077] Laser power 60W, scanning speed 1200mm / s, layer thickness 20μm, track spacing 80μm, strip width 3mm, interlayer rotation angle 67°; argon protection throughout the process, oxygen in the forming chamber ≤500ppm.
[0078] S4 Forming and Sampling
[0079] The printed volume is 12mm×12mm×12mm. The cell sample block was removed by wire EDM, and there were no macroscopic cracks on the surface.
[0080] S5 Multidimensional Test
[0081] a) Surface roughness
[0082] A laser confocal microscope (50× objective lens) was used to select the flat area of the Diamond saddle point. The measured values were Ra=2.8μm, Rz=6.9μm, and Sa=5.5μm, all of which meet the requirements of Ra≤5μm, Rz≤10μm, and Sa≤10μm of claim 3 in the abstract.
[0083] b) Industrial CT Scanning
[0084] After μCT (3μm voxel) scanning, registration is performed using STL as the reference. The minimum wall thickness is 0.342mm, and the absolute deviation from the design value of 0.35mm is -8μm. A positive or negative deviation of ≤20μm is acceptable.
[0085] c) Metallographic density
[0086] Hot mounting → progressive grinding to 2500# → 3μm and 1μm diamond polishing → 0.05μm alumina suspension final polishing; ImageJ binarization of 12 500× images, average density 99.94%, no >3μm unfused or cracked areas observed.
[0087] d) Compression performance
[0088] The sample is 15mm high and 10mm in diameter (height-to-diameter ratio 1.5:1). The parallelism of the upper and lower end faces is 0.006mm after grinding. The displacement of the electronic universal testing machine is controlled at 0.8mm / min. Measured at room temperature:
[0089] Elastic modulus E = 114 GPa, yield strength σ y =1080MPa, peak stress 1250MPa, densification strain ε dens =0.38, with a deviation of <4% from the sampling data of the same process integral heat exchanger.
[0090] S6 Machine Learning Modeling
[0091] Fifty sets of small cell data with different powers (50-70W), speeds (1000-1400mm / s), and layer thicknesses (10-30μm) were used as input, with Ra and t as the data. min ρ rel σ y For the output, a gradient boosting regression tree (GBRT) was used, and 10-fold cross-validation yielded R² = 0.89, which meets the requirement of R² ≥ 0.85 in claim 7 of the abstract. The model predicted the optimal window: power 61W, speed 1230mm / s, layer thickness 20μm, with an error of <2% compared to the actual measurements in this embodiment.
[0092] S7 Performance Verification
[0093] The entire 160mm×180mm×210mm Diamond-Ti65 heat exchanger was reprinted according to the optimized parameters given by the model. Industrial CT sampling inspection showed an average wall thickness deviation of 6μm and a density of 99.2%. It was sealed and pressure held for 15 minutes at 0.5MPa without leakage, proving that the "representative cell sampling-evaluation-prediction" process proposed in this embodiment can efficiently guide the precision additive manufacturing and quality assurance of complex titanium alloy microchannel heat exchangers.
Claims
1. A method for evaluating the forming quality of a precision additive manufacturing microchannel heat exchanger, characterized in that, Includes the following steps: S1 obtains the original three-dimensional STL model of the heat exchanger to be manufactured, the model containing periodically arranged three-period minimal surface (TPMS) cell structures. S2 extracts a representative circulating single cell from the original STL model, adds a simple thin-walled lug with the same characteristic wall thickness to the side of the cell and merges them into a part, and exports it as a local test piece model. The representative cell is consistent with the overall structure of the heat exchanger in terms of topology, wall thickness distribution, porosity and curvature characteristics. S3 uses laser powder bed melting (LPBF) precision additive manufacturing equipment to prepare the local test piece model, obtain metallographic cell samples, and then performs wire cutting separation on the samples; S4 performs multi-dimensional performance tests on the metallographic cell sample, including: a. Surface roughness test: Non-contact scanning of the original cellular surface is performed using a laser confocal microscope or white light interferometer to obtain roughness parameters such as Ra, Rz, and Sa, which are used to evaluate the quality of the formed surface; b. Industrial CT scanning: High-resolution X-ray three-dimensional imaging of the original cell sample is performed to extract information on actual wall thickness distribution, geometric deviation and internal defects, which is used to evaluate the accuracy of thin-wall forming; c. Metallographic sample preparation and observation: After the cell sample is mounted, ground and polished, the cross section is observed using an optical microscope or scanning electron microscope to measure the minimum wall thickness, average wall thickness and local density, and to identify defects such as incomplete fusion, porosity and cracks. d. Mechanical performance testing: Quasi-static compression tests are conducted on cell samples to obtain elastic modulus, yield strength, peak stress, and energy absorption capacity, which are used to evaluate the structural load-bearing performance; Based on the above test results, S5 uses machine learning methods to establish a correlation model of "process parameters - cell quality - heat exchanger performance" to guide the LPBF process optimization and finished product quality prediction of actual heat exchanger components.
2. The method for evaluating the forming quality of a precision additive manufacturing microchannel heat exchanger according to claim 1, characterized in that: The representative cell is cut off at a location within the thin-walled region inside the heat exchanger, with a contour size of one unit cell period. The cut positions on each face are located on the symmetrical neutral plane of the cell structure to avoid boundary effects interfering with the test results.
3. The method for evaluating the forming quality of a precision additive manufacturing microchannel heat exchanger according to claim 1, characterized in that: The surface roughness test area should be selected in a relatively flat area of the metallographic TPMS cell surface to facilitate subsequent leveling. The roughness parameters measured under a 50x laser confocal microscope objective should meet the following requirements: Ra≤5μm, Rz≤10μm, Sa≤10μm.
4. The method for evaluating the forming quality of a precision additive manufacturing microchannel heat exchanger according to claim 1, characterized in that: The industrial CT scan has a resolution of ≤5μm, adopts a local magnification scanning mode, and performs geometric registration based on the STL model. The minimum wall thickness positive and negative deviation is ≤20μm, which is quantitatively output by deviation chromatography.
5. The method for evaluating the forming quality of a precision additive manufacturing microchannel heat exchanger according to claim 1, characterized in that: The metallographic cell samples were cold-mounted or hot-mounted, and then polished stepwise (using fine sandpaper) using diamond polishing paste with a particle size of 3μm and 1μm, respectively. Finally, they were polished with a polishing liquid with a particle size of 0.25μm to obtain mirror-finish metallographic samples. Cell sample observation was performed using image processing software such as ImageJ for image binarization. The density was calculated as (number of solid pixels / total number of pixels) × 100%. There should be no obvious unfused areas or macroscopic cracks in the solid area, and the density should be ≥99.5%.
6. The method for evaluating the forming quality of a precision additive manufacturing microchannel heat exchanger according to claim 1, characterized in that: The compression test adopts a displacement control mode with a loading rate of 0.5-1 mm / min. The height-to-diameter ratio of the sample is 1:1 to 1.5:1, and the upper and lower end faces are ground flat to ensure parallelism ≤0.01 mm. The stress-strain curve is used to extract the platform stress and densification strain.
7. The method for evaluating the forming quality of a precision additive manufacturing microchannel heat exchanger according to claim 1, characterized in that: The "process parameters-cell mass-heat exchanger performance" correlation model is constructed using machine learning algorithms. Input variables include laser power, scanning speed, layer thickness, and channel spacing, while output variables include surface roughness Ra, Rz, Sa, and minimum wall thickness t. min packing density ρ rel Compressive strength σ y After the model is trained, it is used to predict the overall heat exchanger performance under different processes, enabling rapid selection of process windows.
8. The method for evaluating the forming quality of a precision additive manufacturing microchannel heat exchanger according to claim 1, characterized in that: The TPMS structure is of the Gyroid, Diamond, or I-WPSchwarz-P type, with a cell size of 5-20 mm and a cell characteristic wall thickness of 0.1-0.8 mm.
9. The method for evaluating the forming quality of a precision additive manufacturing microchannel heat exchanger according to claim 1, characterized in that: The laser powder bed melting (LPBF) raw materials for manufacturing the microchannel heat exchanger are selected from one of titanium alloys, aluminum alloys, copper alloys, stainless steel or nickel-based high-temperature alloys, and the material is a spherical pre-alloyed powder with excellent spreadability and density during forming.