Metal powder defect classification and quantitative characterization method and system based on geometric morphological parameters

Abstract

The present invention provides a method and system for classifying and quantifying metal powder defects based on geometric morphological parameters, including the following steps: Step 1: Obtaining a metal powder sample and preparing a sample; Step 2: Observing the powder to obtain relevant image data; Step 3: Using a graphic analysis tool to identify, segment, and extract the features of each metal powder to obtain the relevant geometric parameters of the metal powder; Step 4: Extracting a preset number of metal powders as samples based on the metal powder category and training a machine learning classification algorithm; Step 5: Preparing a corresponding sample, obtaining all the geometric parameters of the metal powder, and using the trained machine learning model to classify and quantify the powder morphology to obtain a quantitative map of the metal powder morphology distribution; Step 6: Assessing whether the powder quality and yield meet requirements. This method can more accurately assess metal powder quality, ensure the quality of metal powders used in additive manufacturing, and improve the service performance of additively manufactured components.

Description

Technical Field

[0001] The present invention relates to the technical field of metal material characterization, and in particular to a method and system for classifying and quantifying metal powder defects based on geometric morphological parameters. Background Art

[0002] Metal powder, as an important consumable in additive manufacturing, plays a vital role in the development and application of additive manufacturing. At present, the main preparation process for powder is atomization powder making process, including vacuum melting atomization powder making process, plasma atomization process, electro-induction atomization, rotating electrode atomization process, and water atomization process. These powder making processes provide a large amount of metal powder for additive manufacturing and promote the development of additive manufacturing.

[0003] The morphology of metal powders can significantly impact the performance of additively manufactured parts, primarily by introducing defects such as voids. Currently, most powders produced by powder-making processes have certain defects, particularly in the most widely used aerosol powder-making process. Due to its high yield and low cost, this process accounts for 50% of all powders produced. Powder defects produced by this process can be categorized as hollow powder, satellite powder, and irregularly shaped powder. These powder defects affect the bulk density and flowability of metal powders on a macroscopic level and the generation of void defects in additive manufacturing on a microscopic level. Current methods for characterizing metal powders rely primarily on macroscopic measurements of bulk density, compacted density, and flowability. Microscopically, these methods rely on scanning electron microscopy, metallographic cross-section observations, and CT scans. These methods can only qualitatively or semi-quantitatively determine the quality of metal powders. They are unable to quantify powder quality from the perspective of powder morphology, nor can they provide optimization recommendations for the powder-making process based on this perspective, as different powder defects have their own physical mechanisms.

[0004] Metal powders prepared by various powder making processes have various defects, which will seriously affect the additive manufacturing process. Current powder characterization methods cannot quantify these metal powder defects from the metal powder morphology. Since the physical mechanisms of defects in different metal powders are different, quantifying these defects can optimize the powder making process and powder making equipment, which is of great significance for improving the performance of additive manufacturing parts.

[0005] Patent document CN115078423A (application number: CN202210729483.3) discloses an alloy powder internal pore defect sample, and its characterization method, device, medium and equipment belong to the field of alloy powder technology. The sample includes a non-metallic container, alloy powder and solidified material, the alloy powder is spherical, the alloy powder is contained in the container, and the solidified material is filled between the alloy powders in the container; the alloy powder has a first color (contrast) under X-ray imaging conditions, the solidified material has a second color under X-ray imaging conditions, and the alloy powder internal pore defects have a third color under X-ray imaging conditions, wherein the first color, second color and third color can be clearly distinguished. However, this patent cannot solve the current technical problems. Summary of the Invention

[0006] In view of the defects in the prior art, the purpose of the present invention is to provide a method and system for classifying and quantifying metal powder defects based on geometric morphological parameters.

[0007] The method for classifying and quantifying metal powder defects based on geometric morphological parameters provided by the present invention includes:

[0008] Step 1: Obtain a metal powder sample from the powder preparation process equipment and prepare the metal powder to meet the sample size requirements of the CT or SXCT equipment;

[0009] Step 2: Use CT or SXCT equipment to observe the powder, set observation parameters based on the minimum size of the powder, and obtain relevant image data;

[0010] Step 3: Use graphic analysis tools to identify, segment, and extract the features of each metal powder, and obtain the relevant geometric parameters of the metal powder based on the powder geometric morphology characterization parameters and its geometric definition;

[0011] Step 4: Based on the metal powder category, a preset amount of metal powder is extracted as samples to train the machine learning classification algorithm;

[0012] Step 5: Based on the probabilistic sampling method, prepare the corresponding CT or SXCT sample, obtain all the geometric parameters of its metal powder, and use the trained machine learning model to classify and quantify the powder morphology to obtain a quantitative map of the metal powder morphology distribution;

[0013] Step 6: Based on the metal powder morphology distribution quantitative diagram, assess whether the powder quality and output meet the requirements. If not, optimize and adjust the powder making process and equipment, and return to step 5 to continue.

[0014] Preferably, an atomization powder making process is adopted, including water atomization, gas atomization, vacuum induction gas atomization, plasma atomization and rotary electrode atomization;

[0015] The metal powder includes powders containing metal elements used in additive manufacturing, including iron alloys, aluminum alloys, titanium alloys, copper alloys, and various metal-based composite materials;

[0016] The CT or SXCT equipment is any equipment that can perform three-dimensional imaging analysis on a sample, including CT using various light sources;

[0017] According to the sample preparation requirements of CT or SXCT equipment, obtain a sufficient number of samples to build metal powder 3D model information.

[0018] Preferably, the graphic analysis tool includes commercial image processing software, open source image processing software, and related codes based on image processing algorithms;

[0019] The powder geometric morphology characterization parameters include all geometric parameters of the reaction metal powder, including Feret diameter, sphericity, porosity, roughness, aspect ratio, elongation, ellipsoidality, volume and surface area;

[0020] The metal powder categories include normal powder, special-shaped powder, satellite powder, hollow powder, and various combinations of defects;

[0021] The machine learning classification algorithm includes all machine learning classification algorithms, including decision trees, support vector machines, neural networks, Bayesian classifiers and their combination algorithms.

[0022] Preferably, the probabilistic sampling method includes: based on the central limit theorem, after determining the confidence interval and error according to the actual amount of powder, the required number of samples is used to reflect the overall aerosol powder making situation.

[0023] Preferably, the optimization of the powder making process includes adjusting the gas pressure, the temperature of the molten metal and the flow rate; the optimization of the powder making equipment includes adjusting the nozzle structure of the powder making equipment, as well as additional equipment.

[0024] The metal powder defect classification and quantitative characterization system based on geometric morphological parameters provided by the present invention includes:

[0025] Module M1: Obtain metal powder samples from the powder preparation process equipment and prepare the metal powder to meet the sample size requirements of the CT or SXCT equipment;

[0026] Module M2: Use CT or SXCT equipment to observe the powder, set observation parameters based on the minimum size of the powder, and obtain relevant image data;

[0027] Module M3: Use graphic analysis tools to identify, segment, and extract the features of each metal powder, and obtain the relevant geometric parameters of the metal powder based on the powder geometric morphology characterization parameters and their geometric definitions;

[0028] Module M4: Extract a preset amount of metal powder as samples based on the metal powder category to train the machine learning classification algorithm;

[0029] Module M5: Based on the probabilistic sampling method, prepare the corresponding CT or SXCT specimens, obtain all the geometric parameters of the metal powder, and use the trained machine learning model to classify and quantify the powder morphology to obtain a quantitative map of the metal powder morphology distribution;

[0030] Module M6: Based on the quantitative diagram of metal powder morphology distribution, assess whether the powder quality and output meet the requirements. If it does not meet the production requirements, optimize and adjust the powder making process and powder making equipment, triggering module M5 to continue running.

[0031] Preferably, an atomization powder making process is adopted, including water atomization, gas atomization, vacuum induction gas atomization, plasma atomization and rotary electrode atomization;

[0032] The metal powder includes powders containing metal elements used in additive manufacturing, including iron alloys, aluminum alloys, titanium alloys, copper alloys, and various metal-based composite materials;

[0033] The CT or SXCT equipment is any equipment that can perform three-dimensional imaging analysis on a sample, including CT using various light sources;

[0034] According to the sample preparation requirements of CT or SXCT equipment, obtain a sufficient number of samples to build metal powder 3D model information.

[0035] Preferably, the graphic analysis tool includes commercial image processing software, open source image processing software, and related codes based on image processing algorithms;

[0036] The powder geometric morphology characterization parameters include all geometric parameters of the reaction metal powder, including Feret diameter, sphericity, porosity, roughness, aspect ratio, elongation, ellipsoidality, volume and surface area;

[0037] The metal powder categories include normal powder, special-shaped powder, satellite powder, hollow powder, and various combinations of defects;

[0038] The machine learning classification algorithm includes all machine learning classification algorithms, including decision trees, support vector machines, neural networks, Bayesian classifiers and their combination algorithms.

[0039] Preferably, the probabilistic sampling method includes: based on the central limit theorem, after determining the confidence interval and error according to the actual amount of powder, the required number of samples is used to reflect the overall aerosol powder making situation.

[0040] Preferably, the optimization of the powder making process includes adjusting the gas pressure, the temperature of the molten metal and the flow rate; the optimization of the powder making equipment includes adjusting the nozzle structure of the powder making equipment, as well as additional equipment.

[0041] Compared with the prior art, the present invention has the following beneficial effects:

[0042] (1) The present invention establishes a three-dimensional model of metal powder through three-dimensional characterization, analyzes and quantifies the morphology of metal powder, and thus can more accurately evaluate the quality of metal powder, thereby ensuring the quality of metal powder for additive manufacturing and improving the service performance of additively manufactured parts;

[0043] (2) The present invention can obtain the defects of metal powder and their proportion by quantitatively characterizing the metal powder morphology. The physical mechanisms of defects in different metal powders are different. By analyzing the quantitative parameters of metal powder morphology, guidance can be provided for optimizing the powder making process parameters and the powder making equipment, thereby improving the powder yield and quality of metal powder. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Other features, objects and advantages of the present invention will become more apparent upon reading the detailed description of non-limiting embodiments with reference to the following drawings:

[0045] Figure 1 (a) to (c) are three-dimensional reconstructions of metal powder;

[0046] Figure 2 Classify metal powder defects;

[0047] Figure 3a and Figure 3b The metal powder morphology classification distribution diagram before and after process equipment optimization;

[0048] Figure 4 This is a flow chart of a method for quantitative characterization of metal powder defects based on geometric morphological parameters. DETAILED DESCRIPTION

[0049] The present invention will be described in detail below with reference to specific embodiments. The following examples will help those skilled in the art to further understand the present invention, but are not intended to limit the present invention in any form. It should be noted that, for those skilled in the art, several changes and improvements can be made without departing from the scope of the present invention. These all fall within the scope of protection of the present invention.

[0050] Example 1

[0051] The present invention proposes a method for classifying and quantifying metal powder defects based on geometric morphological parameters, comprising the following steps: obtaining a sufficient number of metal powder samples from a powder preparation process device; preparing the metal powder so that it meets the sample size requirements of a CT or SXCT device; observing the powder using a CT or SXCT method; setting observation parameters based on the minimum size of the powder; performing a three-dimensional reconstruction of the CT or SXCT observation data to obtain a three-dimensional model of the metal powder; using a graphic analysis tool to identify, segment, and extract features of each metal powder; and obtaining relevant geometric parameters of the metal powder based on the powder geometric morphological characterization parameters and their geometric definitions. According to common metal powder classification methods, an appropriate number of corresponding metal powders are extracted as samples to train a machine learning classification algorithm to achieve a satisfactory accuracy rate. Finally, based on a probabilistic sampling method, corresponding CT or SXCT samples are prepared to obtain all the geometric parameters of the metal powders. The trained machine learning model is used to classify and quantify the powder morphology, thereby obtaining a quantitative map of the metal powder morphology distribution. The map is used to guide the powder preparation process and optimize the powder preparation equipment, thereby improving powder quality and output, and providing high-quality and low-cost powder for additive manufacturing.

[0052] The powder making process includes the current atomization powder making process, including water atomization, gas atomization, vacuum induction gas atomization, plasma atomization, rotating electrode atomization and other powder preparation processes.

[0053] The metal powder mentioned includes powders containing metal elements used in additive manufacturing, such as iron alloys, aluminum alloys, titanium alloys, copper alloys, and various metal-based composite materials.

[0054] The CT or SXCT equipment mentioned herein refers to any equipment that can perform three-dimensional imaging analysis on a sample, including CT using various light sources.

[0055] The suitable specimen mentioned herein should be a specimen that can obtain sufficient metal powder three-dimensional model information, and should refer to the sample preparation requirements of CT or SXCT equipment.

[0056] The graphic analysis tools mentioned include commercial image processing software, open source image processing software, and related codes based on image processing algorithms.

[0057] The powder geometric morphology characterization parameters include all geometric parameters of the morphology of the reactive metal powder, including Feret diameter, sphericity, porosity, roughness, aspect ratio, elongation, ellipsoidality, volume, surface area, etc.

[0058] The metal powder classification method mentioned therein is mainly the classification method mentioned in all the literature, which is mainly a classification method based on the powder morphology characteristics, mainly including normal powder, special-shaped powder, satellite powder, hollow powder, and various combinations of defects.

[0059] The machine learning classification algorithms mentioned include all machine learning classification algorithms, including decision trees, support vector machines, neural networks, Bayesian classifiers and other classification algorithms and their combination algorithms.

[0060] The probabilistic sampling method described therein is mainly based on the central limit theorem. After determining the confidence interval and error according to the actual amount of powder, the required number of samples can be obtained to reflect the overall aerosol powder making situation.

[0061] The metal powder morphology distribution quantitative diagram and the relationship between the metal powder morphology classification and other geometric parameters such as particle size, sphericity, aspect ratio, etc. are described.

[0062] The optimization and adjustment of the powder making process and powder making equipment mentioned above refers to the process parameters that affect the powder quality and can be controlled, such as gas pressure, molten metal temperature and flow rate. The optimization of the powder making equipment includes adjusting the nozzle structure of the powder making equipment and additional equipment.

[0063] Example 2

[0064] like Figure 4 Taking TiB2-based AlSi10Mg composite metal material as an example, the above-mentioned method of the present invention is used to characterize and quantify its powder morphology, which specifically includes the following steps:

[0065] Step 1: Obtain a TiB2-based AlSi10Mg composite metal powder sample with a mass of 100 g from the gas atomization powder preparation process equipment, and then prepare it to meet the sample size requirements of the SXCT equipment. As shown in Table 1, the sample size and related parameter settings of SXCT at different resolutions are shown. Here, the parameter settings corresponding to the pixel size of 0.65 μm are selected, and the sample size is

[0066] Table 1: SXCT experimental size requirements and parameter settings

[0067]

[0068] Step 2: Based on the TiB2-based AlSi10Mg composite metal powder sample obtained in step 1, the powder is observed using the SXCT method. The observation parameters are set based on the minimum size of the powder. Here, the minimum size of the powder is 5μm, so the parameter settings used for 0.65μm pixels are used to obtain relevant image data;

[0069] Step 3: For the image data obtained in step 2, use the graphic analysis tool to identify, segment and extract the features of each metal powder, and obtain the relevant geometric parameters of the metal powder based on the powder geometric morphology characterization parameters and its geometric definition, such as Figure 1 As shown in (a), (b) and (c);

[0070] Step 4: According to the common metal powder classification method, the powder form is divided into: hollow powder, bonding powder, special-shaped powder, normal powder, such as Figure 2 As shown, 4000 phase metal powders were extracted as samples to train the machine learning classification algorithm, achieving an accuracy rate of >99%;

[0071] Step 5: Prepare the corresponding SXCT samples based on the probabilistic sampling method. The sampling standards are shown in Table 2:

[0072] Table 2: SXCT experimental size requirements and parameter settings

[0073] Confidence Level 99.99％ Error margin 2.5％ Total aerosolized powder 5.6e+11 Required powder sample 6053

[0074] The prepared sample or multiple samples must contain no less than 6053 metal powders, and all the geometric parameters of the metal powders are obtained. The trained machine learning model is used to classify and quantify the powder morphology, thereby obtaining a quantitative distribution diagram of the metal powder morphology, such as Figure 3a As shown;

[0075] Step 6: According to the quantitative distribution diagram of the metal powder morphology obtained in step 5, it is explained whether the powder output meets the requirements, but the quality does not meet the requirements. There is too much bonding powder. The bonding powder is mainly produced in the primary crushing and secondary crushing of aerosolization. The primary crushing is mainly due to the uneven particle size and powder collision during crushing, and the fibrous crushing produces a large number of droplets and blocks of particle size. In addition to the oscillation crushing, other crushing forms in the secondary crushing will cause uneven particle size and powder collision. Finally, there is eddy current in the flow field, and the small particle powder has strong follow-up ability, so it is easy to flow back to the crushing area and The droplets stick together, so the aerosol powder making process and equipment are adjusted. First of all, the process parameters should reduce the metal liquid flow velocity, increase the superheat of the metal liquid, reduce the metal liquid viscosity, reduce the number of fibers, and shorten the length; secondly, the equipment, the nozzle should be modified, and adjusted according to the position and size of the back pressure area, changed to an elliptical shape, and finally the aerosol flow field distribution should be increased at the vortex confluence. The vortex dissipation structure should be added to break the large vortex into small vortices, dissipate these vortices, and then aerosolize to prepare the powder. Repeat steps five and six, and finally make the number of hollow powder meet the production needs, such as Figure 3b shown.

[0076] Example 3

[0077] The present invention also provides a metal powder defect classification and quantitative characterization system based on geometric morphological parameters. The metal powder defect classification and quantitative characterization system based on geometric morphological parameters can be realized by executing the process steps of the metal powder defect classification and quantitative characterization method based on geometric morphological parameters, that is, those skilled in the art can understand the metal powder defect classification and quantitative characterization method based on geometric morphological parameters as a preferred embodiment of the metal powder defect classification and quantitative characterization system based on geometric morphological parameters.

[0078] The metal powder defect classification and quantification characterization system based on geometric morphological parameters provided by the present invention includes: module M1: obtaining a metal powder sample from a powder preparation process device and preparing the metal powder to meet the sample size requirements required by a CT or SXCT device; module M2: observing the powder using a CT or SXCT device, setting observation parameters based on the minimum size of the powder, and obtaining relevant image data; module M3: using a graphic analysis tool to identify, segment, and extract the features of each metal powder, and obtaining relevant geometric parameters of the metal powder based on the powder geometric morphological characterization parameters and their geometric definitions; module M4: extracting a preset amount of metal powder as a sample based on the metal powder category to train a machine learning classification algorithm; module M5: preparing a corresponding CT or SXCT sample based on a probabilistic sampling method, obtaining all the geometric parameters of the metal powder, and using a trained machine learning model to classify and quantify the powder morphology, thereby obtaining a metal powder morphology distribution quantification map; module M6: assessing whether the powder quality and yield meet the production requirements based on the metal powder morphology distribution quantification map. If it does not meet the production requirements, the powder making process and powder making equipment are optimized and adjusted, triggering module M5 to continue operation.

[0079] Adopt atomization powder making process, including water atomization, gas atomization, vacuum induction gas atomization, plasma atomization and rotary electrode atomization;

[0080] The metal powder includes powders containing metal elements used in additive manufacturing, including iron alloys, aluminum alloys, titanium alloys, copper alloys, and various metal-based composite materials;

[0081] The CT or SXCT equipment is any equipment that can perform three-dimensional imaging analysis on a sample, including CT using various light sources;

[0082] According to the sample preparation requirements of CT or SXCT equipment, obtain a sufficient number of samples to build metal powder 3D model information.

[0083] The graphic analysis tools include commercial image processing software, open source image processing software, and related codes based on image processing algorithms;

[0084] The powder geometric morphology characterization parameters include all geometric parameters of the reaction metal powder, including Feret diameter, sphericity, porosity, roughness, aspect ratio, elongation, ellipsoidality, volume and surface area;

[0085] The metal powder categories include normal powder, special-shaped powder, satellite powder, hollow powder, and various combinations of defects;

[0086] The machine learning classification algorithm includes all machine learning classification algorithms, including decision trees, support vector machines, neural networks, Bayesian classifiers and their combination algorithms.

[0087] The probabilistic sampling method includes: based on the central limit theorem, according to the actual amount of powder, after determining the confidence interval and error, the corresponding required number of samples is used to reflect the overall aerosol powder making situation.

[0088] Optimization of the pulverizing process includes adjusting gas pressure, molten metal temperature and flow rate; optimization of the pulverizing equipment includes adjusting the nozzle structure of the pulverizing equipment and additional equipment.

[0089] Those skilled in the art will appreciate that, in addition to implementing the system, device, and various modules provided by the present invention in purely computer-readable program code, it is entirely possible to implement the same program in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, embedded microcontrollers, and the like by logically programming the method steps. Therefore, the system, device, and various modules provided by the present invention can be considered a hardware component, and the modules included therein for implementing various programs can also be considered structures within the hardware component; the modules for implementing various functions can also be considered both software programs for implementing the method and structures within the hardware component.

[0090] The above describes specific embodiments of the present invention. It should be understood that the present invention is not limited to the specific embodiments described above, and those skilled in the art may make various changes or modifications within the scope of the claims, which do not affect the essence of the present invention. The embodiments of this application and the features in the embodiments may be combined with each other in any manner unless there is a conflict.

Claims

1. A method for quantitative characterization of metal powder defects based on geometric morphological parameters, characterized in that: include: Step 1: Obtain a metal powder sample from the powder preparation process equipment and prepare the metal powder to meet the sample size requirements of the preset equipment; Step 2: Use the preset equipment to observe the powder, set the observation parameters based on the minimum size of the powder, and obtain relevant image data; Step 3: Use graphic analysis tools to identify, segment, and extract the features of each metal powder, and obtain the relevant geometric parameters of the metal powder based on the powder geometric morphology characterization parameters and its geometric definition; Step 4: Based on the metal powder category, a preset amount of metal powder is extracted as samples to train the machine learning classification algorithm; Step 5: Based on the probabilistic sampling method, prepare the corresponding sample, obtain all the geometric parameters of its metal powder, and use the trained machine learning model to classify and quantify the powder morphology to obtain a quantitative distribution map of the metal powder morphology; Step 6: Based on the metal powder morphology distribution quantitative diagram, assess whether the powder quality and output meet the requirements. If not, optimize and adjust the powder making process and equipment, and return to step 5 to continue.

2. The method for classifying and quantifying metal powder defects based on geometric morphological parameters according to claim 1 is characterized in that: Adopt atomization powder making process, including water atomization, gas atomization, vacuum induction gas atomization, plasma atomization and rotary electrode atomization; The metal powder includes powders containing metal elements used in additive manufacturing, including iron alloys, aluminum alloys, titanium alloys, copper alloys, and various metal-based composite materials; The preset equipment is all equipment that can perform three-dimensional imaging analysis on the sample, including CT using various light sources; According to the sample preparation requirements of the preset equipment, obtain a sufficient number of samples to build metal powder three-dimensional model information.

3. The method for classifying and quantifying metal powder defects based on geometric morphological parameters according to claim 1 is characterized in that: The graphic analysis tools include commercial image processing software, open source image processing software, and related codes based on image processing algorithms; The powder geometric morphology characterization parameters include all geometric parameters of the reaction metal powder, including Feret diameter, sphericity, porosity, roughness, aspect ratio, elongation, ellipsoidality, volume and surface area; The metal powder categories include normal powder, special-shaped powder, satellite powder, hollow powder, and various combinations of defects; The machine learning classification algorithm includes all machine learning classification algorithms, including decision trees, support vector machines, neural networks, Bayesian classifiers and their combination algorithms.

4. The method for classifying and quantifying metal powder defects based on geometric morphological parameters according to claim 1, characterized in that: The probabilistic sampling method includes: based on the central limit theorem, according to the actual amount of powder, after determining the confidence interval and error, the corresponding required number of samples is used to reflect the overall aerosol powder making situation.

5. The method for classifying and quantifying metal powder defects based on geometric morphological parameters according to claim 1, characterized in that: Optimization of the pulverizing process includes adjusting gas pressure, molten metal temperature and flow rate; optimization of the pulverizing equipment includes adjusting the nozzle structure of the pulverizing equipment and additional equipment.

6. A metal powder defect classification and quantitative characterization system based on geometric morphological parameters, characterized in that: include: Module M1: Obtain metal powder samples from the powder preparation process equipment and prepare the metal powder to meet the sample size requirements of the preset equipment; Module M2: Observe the powder using a preset device, set observation parameters based on the minimum size of the powder, and obtain relevant image data; Module M3: Use graphic analysis tools to identify, segment, and extract the features of each metal powder, and obtain the relevant geometric parameters of the metal powder based on the powder geometric morphology characterization parameters and their geometric definitions; Module M4: Extract a preset amount of metal powder as samples based on the metal powder category to train the machine learning classification algorithm; Module M5: Based on the probabilistic sampling method, prepare the corresponding sample, obtain all the geometric parameters of the metal powder, and use the trained machine learning model to classify and quantify the powder morphology to obtain a quantitative distribution map of the metal powder morphology; Module M6: Based on the quantitative diagram of metal powder morphology distribution, assess whether the powder quality and output meet the requirements. If it does not meet the production requirements, optimize and adjust the powder making process and powder making equipment, triggering module M5 to continue running.

7. The metal powder defect classification and quantification characterization system based on geometric morphological parameters according to claim 6 is characterized in that: Adopt atomization powder making process, including water atomization, gas atomization, vacuum induction gas atomization, plasma atomization and rotary electrode atomization; The metal powder includes powders containing metal elements used in additive manufacturing, including iron alloys, aluminum alloys, titanium alloys, copper alloys, and various metal-based composite materials; The preset equipment is all equipment that can perform three-dimensional imaging analysis on the sample, including CT using various light sources; According to the sample preparation requirements of the preset equipment, obtain a sufficient number of samples to build metal powder three-dimensional model information.

8. The metal powder defect classification and quantification characterization system based on geometric morphological parameters according to claim 6 is characterized in that: The graphic analysis tools include commercial image processing software, open source image processing software, and related codes based on image processing algorithms; The powder geometric morphology characterization parameters include all geometric parameters of the reaction metal powder, including Feret diameter, sphericity, porosity, roughness, aspect ratio, elongation, ellipsoidality, volume and surface area; The metal powder categories include normal powder, special-shaped powder, satellite powder, hollow powder, and various combinations of defects; The machine learning classification algorithm includes all machine learning classification algorithms, including decision trees, support vector machines, neural networks, Bayesian classifiers and their combination algorithms.

9. The metal powder defect classification and quantification characterization system based on geometric morphological parameters according to claim 6, characterized in that: The probabilistic sampling method includes: based on the central limit theorem, according to the actual amount of powder, after determining the confidence interval and error, the corresponding required number of samples is used to reflect the overall aerosol powder making situation.

10. The metal powder defect classification and quantitative characterization system based on geometric morphological parameters according to claim 6, characterized in that: Optimization of the pulverizing process includes adjusting gas pressure, molten metal temperature and flow rate; optimization of the pulverizing equipment includes adjusting the nozzle structure of the pulverizing equipment and additional equipment.

Images