Method for analyzing properties of metal powder in additive manufacturing process

Abstract

The application relates to the technical field of additive manufacturing, and discloses a metal powder characteristic analysis method in an additive manufacturing process. The method comprises the following steps: acquiring real-time image data and environmental parameter data of the metal powder; pre-processing the real-time image data to obtain a standardized powder particle image; determining a particle size correlation factor of adjacent regions according to pixel distribution of different regions of the image; combining the correlation factor with a change trend of the environmental parameter data to calculate a particle distribution fluctuation index; determining abnormal probability of each physical characteristic dimension of the metal powder based on the index and a preset powder characteristic database; generating a comprehensive evaluation result of powder characteristics according to abnormal probability distribution of all dimensions; and adjusting additive manufacturing control parameters according to the comprehensive evaluation result to complete powder characteristic analysis. The method can realize real-time and comprehensive analysis of powder characteristics, improve analysis accuracy and reliability, and help precise control of the additive manufacturing process.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of additive manufacturing, in particular to a method for analyzing the characteristics of metal powder in an additive manufacturing process. BACKGROUND

[0002] In the development process of additive manufacturing technology, as the core raw material, the characteristics of metal powder are directly related to the quality and performance of the final formed component. With the expansion of additive manufacturing application scenarios to high-precision fields such as aerospace and medical implantation, the control of physical properties such as particle size distribution, sphericity and fluidity of metal powder is becoming increasingly strict. At present, the analysis of the characteristics of metal powder in the industry mainly relies on offline detection methods, that is, by sampling after the additive manufacturing process, using laser particle size analyzers, scanning electron microscopes and other equipment for detection. Such methods have obvious limitations. On the one hand, sampling detection cannot reflect the dynamic changes of the characteristics of metal powder in the manufacturing process in real time. When the detection finds that the characteristics of the powder are abnormal, a large number of unqualified components have been produced, resulting in serious waste of material and time cost. On the other hand, offline detection can only obtain the characteristics data of local powder samples, which is difficult to fully represent the real state of the entire powder system, and may result in a situation where the detection results do not match the actual powder characteristics.

[0003] The existing method for analyzing the characteristics of metal powder often ignores the influence of environmental parameters on the characteristics of the powder. In the additive manufacturing process, environmental factors such as temperature, humidity and inert gas concentration continuously act on the metal powder, which may cause problems such as powder particle agglomeration and oxidation, thereby changing the key characteristics of the powder such as particle size distribution and fluidity. The traditional analysis method only focuses on the image data of the powder itself and does not include the change of environmental parameters in the analysis system, which leads to inaccurate judgment of the reasons for the abnormal characteristics of the powder and cannot provide a comprehensive basis for subsequent control parameter adjustment. In addition, the existing analysis method for evaluating the characteristics of the powder is mainly limited to a single dimension, and lacks comprehensive consideration of multiple physical characteristic dimensions, which makes it difficult to form a comprehensive evaluation result of the powder characteristics, and makes it difficult for the operator to clearly understand the overall state of the powder, affecting the precise control of the additive manufacturing process. SUMMARY

[0004] The purpose of the present application is to provide a method for analyzing the characteristics of metal powder in an additive manufacturing process to solve the problems raised in the background.

[0005] To achieve the above-mentioned purpose, the present application provides a method for analyzing the characteristics of metal powder in an additive manufacturing process, which comprises:

[0006] obtaining real-time image data and environmental parameter data of the metal powder in the additive manufacturing process;

[0007] preprocessing the real-time image data to obtain a standardized powder particle image;

[0008] determining a particle size correlation factor of adjacent regions according to pixel distribution of different regions in the standardized powder particle image;

[0009] calculating a particle distribution fluctuation index according to the particle size correlation factor of adjacent regions and the change trend of the environmental parameter data;

[0010] determining an abnormal probability of the metal powder in each physical property dimension based on the particle distribution fluctuation index and a preset powder property database;

[0011] generating a comprehensive evaluation result of the metal powder properties according to the abnormal probability distribution in all physical property dimensions;

[0012] adjusting the additive manufacturing control parameters according to the comprehensive evaluation result of the metal powder properties, and completing the metal powder property analysis.

[0013] Preferably, the real-time image data and environmental parameter data of the metal powder in the additive manufacturing process are obtained, comprising:

[0014] presetting a sampling frequency of the image acquisition device and a monitoring period of the environmental sensor;

[0015] capturing continuous frame image data of the metal powder by using the image acquisition device during the additive manufacturing process;

[0016] synchronously recording temperature and humidity parameter data by using the environmental sensor;

[0017] integrating the continuous frame image data and the temperature and humidity parameter data into the real-time image data and the environmental parameter data.

[0018] Preferably, the real-time image data is preprocessed to obtain a standardized powder particle image, comprising:

[0019] performing noise filtering and contrast enhancement operations on the real-time image data;

[0020] segmenting the enhanced image into multiple sub-regions;

[0021] performing particle edge detection on each sub-region to generate a standardized powder particle image.

[0022] Preferably, the particle size correlation factor of adjacent regions is determined according to the pixel distribution of different regions in the standardized powder particle image, comprising:

[0023] extracting pixel density values of a first region and a second region in the standardized powder particle image;

[0024] calculating the covariance of the pixel density values of the first region and the pixel density values of the second region in the time sequence;

[0025] Based on the covariance value, a particle size correlation factor between the first region and the second region is determined.

[0026] Preferably, the particle distribution fluctuation index is calculated according to a change trend of the particle size correlation factor of the adjacent region and the environmental parameter data, including:

[0027] A historical mean value of the particle size correlation factor of the adjacent region is obtained;

[0028] A change trend slope of the environmental parameter data is analyzed;

[0029] A product of the historical mean value and the change trend slope is taken as a basic fluctuation amount;

[0030] The basic fluctuation amount is normalized to obtain the particle distribution fluctuation index.

[0031] Preferably, the abnormal probability of the metal powder in each physical property dimension is determined based on the particle distribution fluctuation index and a preset powder property database, including:

[0032] A matching reference range of the particle distribution fluctuation index is retrieved from the preset powder property database;

[0033] A difference amount between the particle distribution fluctuation index and the matching reference range is compared;

[0034] An initial abnormal value in each physical property dimension is calculated according to the size of the difference amount;

[0035] The initial abnormal value is weightedly averaged to obtain the abnormal probability of the metal powder in each physical property dimension.

[0036] Preferably, a comprehensive evaluation result of the metal powder property is generated according to the abnormal probability distribution in all physical property dimensions, including:

[0037] Abnormal probability values in all physical property dimensions are aggregated;

[0038] A variance and a mean value of the aggregated abnormal probability values are calculated;

[0039] The comprehensive evaluation result of the metal powder property is generated based on the variance and the mean value.

[0040] Preferably, the additive manufacturing control parameter is adjusted according to the comprehensive evaluation result of the metal powder property, including:

[0041] A key index in the comprehensive evaluation result of the metal powder property is extracted;

[0042] The key index is compared with a preset threshold to generate a control parameter adjustment amount;

[0043] Layer thickness and scanning speed parameters of the additive manufacturing equipment are updated based on the control parameter adjustment amount.

[0044] Preferably, after the metal powder characteristic analysis is completed, the method further comprises:

[0045] Monitoring the actual execution effect of the updated layer thickness and scanning speed parameters;

[0046] Collecting the execution effect data as feedback input to the acquisition of real-time image data and environmental parameter data.

[0047] Preferably, the method further comprises an iterative optimization step:

[0048] After generating the comprehensive evaluation result of the metal powder characteristics, checking the convergence state of the evaluation result;

[0049] When the convergence state does not meet the preset condition, repeating the calculation step of the particle distribution fluctuation index and the subsequent steps until the convergence state meets the preset condition.

[0050] Compared with the prior art, the method has the following beneficial effects:

[0051] The metal powder characteristic analysis method in the additive manufacturing process breaks the limitation of traditional offline detection that cannot reflect the dynamic changes of powder characteristics in real time by synchronously acquiring real-time image data and environmental parameter data of the metal powder. The real-time image data can continuously capture the information of the morphology and distribution of the powder particles in the manufacturing process, and the environmental parameter data can record the changes of external factors such as temperature and humidity. The combination of the two makes the analysis process closely follow the additive manufacturing process, timely find the abnormal changes of the powder characteristics, avoid the large number of unqualified components caused by lagging detection, and effectively reduce the waste of material and time cost. At the same time, real-time data acquisition does not need to take local samples and can cover the entire powder system. The data obtained can more comprehensively represent the real state of the powder, improving the accuracy and reliability of the analysis results of the powder characteristics.

[0052] In the data analysis link, the method first pre-processes the real-time image data to obtain a standardized powder particle image, and then determines the particle size correlation factor of adjacent regions according to the pixel distribution in different regions of the image. This process eliminates possible interference factors in the image acquisition process, such as uneven lighting and equipment noise, through standardization processing, ensuring the accuracy of subsequent particle size correlation factor calculation. The introduction of the particle size correlation factor of adjacent regions can accurately reflect the correlation of powder particles in spatial distribution, providing a reliable basis for the calculation of the particle distribution fluctuation index. Subsequently, the particle distribution fluctuation index is calculated by combining the particle size correlation factor of adjacent regions and the change trend of environmental parameter data, which incorporates environmental parameter changes into the analysis system, fully considering the influence of external factors on powder characteristics. This makes the particle distribution fluctuation index not only reflect the changes in the particle size distribution of the powder itself, but also reflect the dynamic adjustment of the powder characteristics under the action of environmental factors, thereby more comprehensively revealing the change law of the powder characteristics and providing more abundant information for judging the reasons for abnormal powder characteristics.

[0053] Based on the particle distribution fluctuation index and the preset powder characteristic database, the abnormal probability of the metal powder in each physical characteristic dimension is determined. With a large amount of powder characteristic data and abnormal cases in the preset database, the current powder characteristics corresponding to the abnormal situation can be quickly matched, improving the efficiency and accuracy of abnormal probability calculation. At the same time, the abnormal probability is calculated for each physical characteristic dimension, which can clearly present the state of the powder characteristics in different dimensions and avoid the limitations of traditional single-dimensional evaluation. On this basis, the comprehensive evaluation result is generated according to the abnormal probability distribution of all physical characteristic dimensions, which can integrate the characteristic information of multiple dimensions into a comprehensive and intuitive evaluation conclusion, allowing the operator to clearly understand the overall state of the powder and identify the strengths and weaknesses of the current powder characteristics.

[0054] According to the comprehensive evaluation result, the additive manufacturing control parameters are adjusted, realizing the close connection between powder characteristic analysis and manufacturing process control. Since the comprehensive evaluation result is comprehensive and accurate, the control parameter adjustment scheme based on it is more targeted, which can accurately solve the problem of abnormal powder characteristics, such as adjusting the environmental humidity control parameter to improve the powder state caused by excessive humidity, or optimizing the heating or cooling system parameters for the abnormal particle size distribution of the powder caused by temperature changes. This targeted adjustment method can effectively improve the metal powder characteristics, ensure the stable progress of the additive manufacturing process, and thus improve the quality and performance of the final formed components, meeting the stringent requirements of high-precision fields such as aerospace and medical implants for component quality. In addition, the entire analysis process does not require complex manual operation, and through automated data collection, processing and analysis, the workload of the operator is reduced, and the level of automatic control of the additive manufacturing process is improved, creating favorable conditions for the scale and precision application of additive manufacturing technology. BRIEF DESCRIPTION OF DRAWINGS

[0055] Figure 1 Working principle diagram of the metal powder characteristic analysis method in the additive manufacturing process described in the present application;

[0056] Figure 2 Flow chart of real-time image data and environmental parameter data;

[0057] Figure 3 Flow chart of standardizing powder particle images;

[0058] Figure 4 Flow chart of calculating particle distribution fluctuation index. DETAILED DESCRIPTION

[0059] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0060] Please refer to Figure 1 The present application provides a metal powder characteristic analysis method in an additive manufacturing process, which comprises:

[0061] During the operation of the additive manufacturing equipment, real-time image data reflecting the powder laying state is continuously acquired by the integrated image acquisition device, and at the same time, environmental parameter data such as temperature and humidity is synchronously acquired by the environmental sensor deployed in the manufacturing cabin. These raw data are transmitted to the data processing system for subsequent analysis. The real-time image data is preprocessed, including noise filtering using a digital filtering algorithm and adjusting the contrast using an image enhancement algorithm, to obtain clearer powder morphology information. The preprocessed image is segmented into several regular sub-regions, and then the particle edge detection is performed on each sub-region, and finally the standardized powder particle image with uniform format and scale is generated. Subsequently, the pixel distribution characteristics of different regions in the standardized powder particle image are analyzed, and the statistical correlation of the pixel density values between adjacent regions is calculated to determine the particle size correlation factor of the adjacent regions. The factor is combined with the trend of the environmental parameter data to calculate the particle distribution fluctuation index through a mathematical operation model. The system accesses a preset powder property database which stores the standard particle distribution fluctuation range and corresponding physical property representation of different types of metal powder under various working conditions. By comparing the real-time calculated particle distribution fluctuation index with the matching reference range in the database, the abnormal probability of the metal powder in each physical property dimension (such as flowability, bulk density, particle size distribution, etc.) is calculated. The abnormal probability values in all physical property dimensions are statistically aggregated and analyzed to generate a comprehensive metal powder property evaluation result. Finally, the evaluation result is input into the control system of the additive manufacturing equipment to drive the adjustment of key process parameters such as layer thickness and scanning speed, thereby realizing online optimization and closed-loop control of the manufacturing process.

[0062] Example 1: see Figure 2 When implementing metal powder property analysis in the additive manufacturing process, obtaining high-quality and synchronized real-time image data and environmental parameter data is the basis of the entire analysis process. The implementation of this process relies on the precise configuration of the hardware system and the coordination of the data acquisition strategy. In actual operation, the image acquisition device is selected as a high-frequency industrial line array camera. Its specific sampling frequency is not a fixed value, but is dynamically preset according to the category of the metal powder used and its powder laying characteristics in the specific additive manufacturing equipment. For example, after the powder laying scraper completes a one-way stroke, there is a short period of stagnation, during which the powder bed is in a relatively static state. The camera uses this stagnation period as a trigger signal for image capture, which can effectively avoid motion blur and obtain clear static images of the powder surface. The installation position of the camera is calculated precisely, usually perpendicular to the powder laying plane and located behind the scraper travel path, to ensure that the original state of the powder layer just completed by laying is captured without being disturbed by laser scanning, with minimal interference. Its lighting system uses a specific angle of low-angle ring LED cold light source to maximize the highlighting of the surface texture and contour of the powder particles and reduce the interference caused by mirror reflection.

[0063] The environmental monitoring unit is synchronously started, and a high-precision, low-delay model is selected for the temperature and humidity sensor. The monitoring period of the sensor is hard-synchronized with the sampling frequency of the camera or is soft-synchronized through the central controller, so as to ensure that each image frame corresponds to a set of environmental parameter readings with the same timestamp. The sensor is arranged at a position that is usually above the powder bed in the manufacturing cabin, so as to avoid interference with moving parts and to be able to sense the air state that best represents the actual environment of the powder. The monitoring of temperature parameter data is crucial for identifying the microclimate changes in the cabin caused by process heat accumulation or equipment heat dissipation, and the monitoring of humidity parameter data is directly related to whether the powder is hygroscopic and caked, thereby affecting the flowability and uniformity of the powder.

[0064] After the device starts a building task, the data acquisition process is started, and the image acquisition device captures a high-resolution grayscale image at the end of each powder laying period according to the preset rhythm. These continuous frame image data are stored in sequence and are attached with metadata such as timestamps and layer numbers. Between the physically adjacent continuous frames, the content shows a high degree of correlation, but also records the subtle evolution of the powder bed state with the increase of the layer number, and even the appearance and expansion of abnormalities, such as scratches, splashes or uneven powder laying. Synchronously, the environmental sensor records the temperature reading and humidity reading in the manufacturing cabin at the time corresponding to each frame image in a completely consistent time sequence. These environmental parameter data are also attached with precise timestamps.

[0065] All the collected raw data are sent to a data preprocessing module, which first pairs the image frames with the environmental parameter data according to the timestamps, so as to ensure the strict consistency of the data in the time sequence. Subsequently, the paired data are preliminarily encapsulated to form a structured data packet. The data packet usually contains a unique serial number, a timestamp, a layer number, a whole frame of image pixel data, a temperature floating point number and a humidity floating point number. This standardized encapsulation format enables the subsequent processing program to efficiently parse and call these data. The integrated real-time image data and environmental parameter data stream are continuously transmitted to the central storage unit or directly sent to the real-time analysis engine through a high-speed data bus.

[0066] Example 2: see Figure 3Embodiments thereof begin with the processing of raw real-time image data captured by the image acquisition system. These raw image data inevitably contain various kinds of disturbances, such as shadows and glares due to uneven cabin illumination, slight blurs caused by minor vibrations of the equipment, and electronic noise inherent to the image sensor itself. To address these noises, a digital filtering algorithm is employed, which is capable of effectively smoothing out isolated noise points in the image while preserving the edge detail information of the powder particles to the maximum extent. After noise suppression, the overall contrast of the image can still be suboptimal, especially when the powder material has weak light-reflecting properties or is close in color to the substrate. Therefore, a contrast enhancement operation is performed, which stretches the dynamic range of the image by redistributing the gray scale values of the pixels, so that the gray scale differences between the powder particles and the background, as well as between the particles and the particles, are amplified, highlighting more texture and contour information. The image after these preliminary processes has its visual quality significantly improved, laying a foundation for the subsequent fine analysis.

[0067] The enhanced image is segmented into multiple sub-regions, with a regular grid partitioning strategy treating the entire image as a two-dimensional plane and dividing it into a number of rectangular blocks of uniform size. The size of each sub-region is carefully set, with its area being large enough to contain tens of powder particles to ensure the reliability of the subsequent statistical analysis, while not being too large to obscure the distribution characteristics of the local region. This partitioning approach transforms the complex global image analysis problem into a number of relatively simple and more manageable local region analysis problems. For each independent sub-region, a particle edge detection operation is performed. The purpose of this operation is to accurately identify the boundaries of each powder particle. The edge detection algorithm achieves this goal by calculating the gray scale gradient intensity of each pixel in the image in different directions. Those pixel points with gradient intensity significantly higher than the surrounding region are determined as edge points. By adjusting the sensitivity parameter of the algorithm, the degree of detail of the detected edges can be controlled, ensuring that the main particle contours can be captured while effectively ignoring the minor scratches or noise-induced false edges present in the image. The result of the edge detection is converted into a binary image representation, i.e., each pixel in the image is either black or white, with the white pixel points representing the identified particle edges and the black regions representing the background or other non-edge regions. This series of processing outputs for each sub-region, i.e., a set of binary images of uniform format containing only clear particle edge information, are defined as standardized powder particle images. These images eliminate most of the visual information in the original data that is unrelated to the characteristics of the particles, preparing for the subsequent quantitative calculations.

[0068] After obtaining the normalized powder particle images, the analysis process enters the stage of determining the particle size correlation factor of adjacent regions. This analysis focuses on two specific sub-regions in the image that are adjacent in spatial position, which can be named as the first region and the second region. First, it is necessary to extract quantitative indicators from these normalized images that can represent the physical characteristics of the particles. For any region, the pixel density value is selected as the core characteristic quantity. The calculation of this value is based on the binarized normalized image, that is, the number of all pixel points marked as white edges in the region is counted, and then divided by the total number of pixel points in the region to obtain a ratio. This ratio reflects the overall size distribution and packing density of the powder particles in the region to some extent; generally speaking, under the premise of relatively uniform particle size, the higher the ratio, the more likely it is that the region is dominated by small particles or the particles are more tightly packed, because small particles will increase the total edge length. The calculation results of a single frame of image may be affected by accidental factors, so it is necessary to analyze from the perspective of time series. The system extracts the pixel density values of the first region and the second region at the current time and at a plurality of consecutive time points (corresponding to a plurality of consecutive powder layers) before the current time, thereby forming two data sequences that change over time. These two sequences respectively depict the dynamic evolution process of the powder particle distribution state of the two adjacent regions.

[0069] The covariance between the time series of the pixel density values of the first region and the time series of the pixel density values of the second region is calculated. Covariance is a statistical quantity that measures the correlation between the changes of two variables. Its calculation process involves calculating the average of the product of the deviations of the two sequences from their respective means. A larger positive covariance value indicates that when the pixel density value of the first region is higher than its average (e.g., the particles become finer or more densely packed), the pixel density value of the second region also tends to be higher than its own average, and vice versa. This indicates that the powder particle distribution states of the two adjacent regions exhibit a high positive correlation and synchronous change trend in the time dimension. Conversely, a covariance value close to zero indicates that the changes of the two regions are relatively independent and lack linkage. A negative covariance value means that when one region state rises, the other region state falls, showing a kind of give-and-take relationship. Based on the calculated covariance value, it is mapped to a more interpretable indicator, the particle size correlation factor between the first region and the second region, through a predefined conversion relationship. This factor is a dimensionless scalar value that directly reflects the correlation strength and consistency degree of the two spatially adjacent regions in the powder particle distribution characteristics. The establishment of this factor converts the visual correlation in the image into a quantitative feature that can be utilized by subsequent mathematical models, realizing a key leap from pixel information to process knowledge.

[0070] Example 3: see Figure 4The spatial correlation features extracted from the images are dynamically combined with the process trend reflected by the environmental sensor data to form a comprehensive quantitative indicator. The calculation of this indicator does not rely on snapshot data at a single moment, but is based on time series analysis, aiming to capture the instability of powder behavior driven by environmental changes. Its implementation begins with the invocation of historical data, and the system maintains an updated data buffer that stores a series of adjacent regional particle size correlation factors calculated in the recent period. This factor, as mentioned earlier, quantifies the correlation strength of the powder particle distribution state in two specific spatial neighborhoods in the image. When calculating the current required particle distribution fluctuation indicator, the system extracts historical data within a time window from this buffer, and the length of this window may cover dozens or even hundreds of powder layers in the past to ensure the stability of the statistics. The arithmetic mean of these historical data points is calculated, which is the historical mean of the adjacent regional particle size correlation factor. This historical mean represents the typical correlation level of the specific region to the powder distribution in a relatively stable operation period in the recent period, and it exists as a benchmark or expected value.

[0071] The system also analyzes the trend of environmental parameter data in parallel, with environmental parameters, especially humidity, being the focus due to their direct impact on the physical properties of the powder. The system selects a sequence of humidity parameter data within the same time window as the image data mentioned above, which consists of a series of time-ordered humidity readings. Linear regression analysis is performed on this sequence, aiming to best fit the change of these data points over time with a straight line. The slope of this fitting line is extracted, which is the slope of the change trend of the environmental parameter data. This slope is a signed number, and its absolute value represents the degree of humidity change, while the sign indicates the direction of change (for example, a positive slope indicates that humidity continues to rise, which may be due to minor fluctuations in the environmental control system or the release of moisture in the process).

[0072] The two calculation results mentioned above, the historical mean and the change trend slope, are combined. The combination is to multiply them directly, and the product is defined as an intermediate variable in the calculation of the particle distribution fluctuation indicator, called the basic fluctuation quantity. The calculation logic is based on the physical understanding that the directed change of environmental parameters (represented by the slope) will act as a driving force to amplify or stimulate the inherent distribution instability of the powder system, while this inherent instability tendency is implicitly represented by the average level of its historical correlation (historical mean). Therefore, the product of the two can comprehensively reflect the potential of powder distribution fluctuation triggered or exacerbated by external environmental disturbance.

[0073] This base fluctuation is a raw value with physical dimension and its value range can fluctuate greatly, which cannot be directly compared with the standard range in the preset database. Therefore, it must be normalized to map it to a unified, dimensionless scalar range. The normalization process uses the maximum and minimum value scaling method. The system maintains a reference range based on long-term process data statistics, which defines the maximum and minimum values of the base fluctuation that can occur under normal operating conditions.

[0074] The normalization process is represented by the following formula:

[0075]

[0076] Wherein: represents the final calculated particle distribution fluctuation index, which is a dimensionless value between 0 and 1. represents the calculated base fluctuation, which is the product of the historical mean and the change trend slope. represents the expected lower limit value of the base fluctuation based on long-term statistical data. represents the expected upper limit value of the base fluctuation based on long-term statistical data.

[0077] Through this calculation, the original base fluctuation is converted into a standardized index. The closer the value of this index to 0, the closer the current powder distribution fluctuation potential to the historical minimum level, indicating a very stable state. The closer to 1, the fluctuation potential reaches the highest level within the historical statistical range, indicating that the powder distribution may be becoming abnormally unstable. The final particle distribution fluctuation index is a comprehensive feature quantity that integrates temporal and spatial information and environmental dynamics, providing a core input basis for subsequent knowledge-based powder property anomaly probability calculation. Embodiment 4: Based on the particle distribution fluctuation index and the preset knowledge base, the anomaly probability of the powder in each physical property dimension is determined, and the final comprehensive evaluation result is generated, which is a decision-making process that integrates real-time monitoring data and historical experience knowledge. The implementation of this process relies on a structured powder property database, which is not simply a data set, but a knowledge system that contains the mapping relationship between the particle distribution behavior and the physical properties of various metal powders under different environmental conditions. Its implementation begins with the further interpretation of the real-time calculated particle distribution fluctuation index. The system accesses a preset powder property database. The database is organized in table form, and one of its core contents, see Table 1, stores reference information for specific materials under different environmental conditions.

[0078] Table 1: Powder Property Database Query Table

[0079] .

[0080] The computing process first needs to retrieve the reference range that matches the current working condition from the database. The system takes the metal powder material type currently being processed (e.g. 316L stainless steel) and the average ambient humidity value currently being monitored (e.g. 22.5% RH) as the joint query condition. According to the table content, the system locates to the row of data that corresponds to the material type “316L stainless steel” and the humidity range “20.1-30.0”, and retrieves the corresponding particle distribution fluctuation index reference range, whose lower limit value is 0.20 and upper limit value is 0.40. This reference range represents the interval that the fluctuation index usually falls into when this material is working normally under this humidity environment in the historical data. The system compares the real-time calculated particle distribution fluctuation index with this retrieved matching reference range. The difference amount is not simply the difference value, but the relative degree of the index value deviating from the reference interval median. The reference interval median is the sum of the upper limit and lower limit divided by two. The calculated difference amount is a value that quantifies the degree of deviation from the historical normal benchmark under the current state.

[0081] According to the size of this difference amount, the system calculates the initial abnormal value for each physical property dimension. The initial abnormal value of each dimension is generated by a predefined conversion rule, which is usually designed as a monotonic increasing function of the difference amount. This means that the larger the difference amount, the higher the calculated initial abnormal value of each dimension. These initial abnormal values reflect the original score of the possible abnormal situation of each single dimension under the current fluctuation index. Since different physical properties have different degrees of influence on the final product quality, the next step needs to weight and integrate all the initial abnormal values of the dimensions. The weight coefficients are also obtained from the database table, and for the retrieved record, the flowability weight coefficient, the loose bulk density weight coefficient, and the particle size distribution weight coefficient are read. These weight coefficients are determined by domain experts based on a large number of process experiments and experience, and their sum is 1. The weighted average calculation is to multiply the initial abnormal value of each dimension by its corresponding weight coefficient, and then add all the product results, finally obtaining a set of values between 0 and 1, i.e. the abnormal probability of the metal powder in each physical property dimension. For example, the abnormal probability of flowability may be as high as 0.85, while the abnormal probability of loose bulk density may be 0.60, and the abnormal probability of particle size distribution may be 0.55. These probability values represent the size of the possibility of abnormality in each dimension, respectively.

[0082] After obtaining the anomaly probability values for all physical property dimensions, the system enters the phase of generating the comprehensive assessment result of metal powder properties. This process first aggregates the anomaly probability values of all dimensions to form a probability set. Statistical analysis is performed on this set to calculate its mean and variance. The mean reflects the average level of anomaly probability across all dimensions, providing a general measure of the overall state of the powder. The variance reflects the degree of dispersion among the anomaly probability values of different dimensions; a higher variance indicates that some dimensions are severely abnormal while others are relatively normal, possibly pointing to a process problem; a lower variance indicates that all dimensions are at a similar level of abnormality, possibly implying a systematic problem. The generation of the comprehensive assessment result is based on the calculated mean and variance, achieved through a logical judgment rule or a comprehensive scoring model. The system can define multiple assessment levels, such as "normal", "attention", and "abnormal". Each level corresponds to a combined range of mean and variance.

[0083] Example 5: Adjusting additive manufacturing control parameters based on the comprehensive assessment result of metal powder properties and completing subsequent monitoring and iterative optimization is the core link of intelligent closed-loop control. The implementation of this process converts the conclusive output obtained in the previous analysis steps into specific device instructions and tracks and feeds back the instruction execution effect. The implementation process begins with the analysis of the generated comprehensive assessment result of metal powder properties. This comprehensive assessment result is a structured data object that contains the overall rating of the powder state and the anomaly probability values of each physical property dimension. The system extracts pre-defined key indicators from this result, which are usually those factors that have the most significant impact on process stability. For example, if the comprehensive assessment indicates that the overall state is "abnormal" and the anomaly probability value of flowability is significantly higher than that of other dimensions, then the flowability anomaly probability will be extracted as a key indicator.

[0084] The system compares these extracted key indicators with preset threshold values stored in the system's control rule base, which are critical values pre-set based on a large number of process experiments, material property knowledge and expert experience. The comparison operation will produce a clear logical judgment. For example, the rule base may be set: if the abnormality probability value of fluidity exceeds 0.75, it is determined that the process parameters need to be adjusted to compensate for the decline in fluidity. Based on this comparison result, the system will generate specific control parameter adjustment amount according to the built-in control strategy model. The model maps the numerical value of the abnormality probability to the magnitude of the adjustment amount. For example, the higher the fluidity abnormality probability value, the larger the layer thickness reduction amount calculated and generated. The adjustment amount is an instruction with a sign and a specific value, which clearly indicates which parameter needs to be changed in which direction (increase or decrease) and to what extent. The system sends the generated control parameter adjustment amount to the motion control system of the additive manufacturing equipment through the application program interface provided by the equipment. After receiving the instruction, the system dynamically updates the parameter set of the construction task being executed, mainly updating the layer thickness and scanning speed parameters. The adjustment of the layer thickness directly affects the laying thickness of each layer of powder, while the adjustment of the scanning speed changes the speed of laser energy input, both of which work together to adapt to the actual properties of the current powder, trying to make the manufacturing process return to a stable state.

[0085] After completing the parameter adjustment, the system immediately starts monitoring the actual execution effect of the new parameter set. This monitoring is achieved by continuously running the analysis method. During the subsequent powder laying process, the image acquisition and environmental sensing system continues to work, capturing new image data and environmental parameter data. These new data are fed into the analysis process to recalculate the particle distribution fluctuation indicators, the abnormality probability of each dimension, and generate new comprehensive evaluation results of the metal powder properties. The core information of these new results, especially the trend of the probability value change of the physical property dimension that appeared abnormal before, is collected as a feedback signal. These execution effect data are input into the front end of the system, i.e. the acquisition step of real-time image data and environmental parameter data, thereby starting a new analysis-adjustment cycle. This makes the entire system form a closed-loop control loop based on a negative feedback mechanism, which can continuously adapt to the dynamic changes of the powder state in the manufacturing process.

[0086] After each generation of the comprehensive assessment result of the metal powder properties, the system automatically checks the convergence state of the assessment result. The judgment of the convergence state is based on a set of pre-set logical conditions, for example, it is required that in the last three analysis cycles, the comprehensive assessment level remains "normal" and the anomaly probability values of all key physical properties show a downward trend and are lower than zero point three. When the system detects that the current assessment result fails to meet these pre-set convergence conditions, for example, the anomaly probability is still fluctuating or is high, it will automatically trigger a new round of complete analysis. This means that the system will use the latest collected data to re-execute all subsequent steps from calculating the particle distribution fluctuation index, re-calculate the anomaly probability, generate a new comprehensive assessment, and decide whether further parameter adjustment is needed. This process will continue to circulate until the system judges that the assessment result has met the pre-set convergence conditions, indicating that the powder properties and process parameters have reached a satisfactory balance state, and the manufacturing process tends to be stable. This iterative mechanism ensures that the system can continuously track and optimize the process state until the target is reached.

[0087] It should be noted that the relational terms herein such as first and second and the like are used solely to distinguish one entity or action from another, without necessarily requiring or implying any such actual relationship or order between or among such entities or actions. Moreover, the terms "comprises", "comprising", or any other variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can include other elements not expressly listed or inherent to such process, method, article, or apparatus.

[0088] Although embodiments of the present application have been shown and described, it is to be understood that various modifications, substitutions, replacements and changes can be made to these embodiments without departing from the principles and spirit of the present application, and the scope of the present application is defined by the appended claims and their equivalents.

Claims

1. A method for analyzing properties of a metal powder in an additive manufacturing process, characterized by, The method comprises the following steps: Obtain real-time image data and environmental parameter data of metal powder in the additive manufacturing process; Preprocess the real-time image data to obtain a standardized powder particle image; Determine the particle size correlation factor between adjacent regions according to the pixel distribution in the standardized powder particle image; Calculate the particle distribution fluctuation index according to the change trend of the particle size correlation factor between adjacent regions and the environmental parameter data; Determine the abnormal probability of the metal powder in each physical property dimension based on the particle distribution fluctuation index and the preset powder property database; Generate a comprehensive evaluation result of the metal powder properties according to the abnormal probability distribution in all physical property dimensions; Adjust the additive manufacturing control parameters according to the comprehensive evaluation result of the metal powder properties, and complete the metal powder property analysis.

2. The method of claim 1, wherein the metal powder is used in an additive manufacturing process. The method comprises the following steps: Pre-set the sampling frequency of the image acquisition device and the monitoring period of the environmental sensor; During the additive manufacturing process, use the image acquisition device to capture continuous frame image data of the metal powder; Use the environmental sensor to record temperature and humidity parameter data synchronously; Integrate the continuous frame image data and the temperature and humidity parameter data into real-time image data and environmental parameter data.

3. The method of claim 1, wherein the metal powder is used in an additive manufacturing process. The method comprises the following steps: Perform noise filtering and contrast enhancement operations on the real-time image data; Segment the enhanced image into multiple sub-regions; Perform particle edge detection on each sub-region to generate a standardized powder particle image.

4. The method of claim 3, wherein the metal powder is a metal powder used in an additive manufacturing process. The method comprises the following steps: Extract the pixel density values of the first region and the second region in the standardized powder particle image; Calculate the covariance of the pixel density values of the first region and the second region in the time series; Determine the particle size correlation factor between the first region and the second region based on the covariance value.

5. The method of claim 1, wherein, The method comprises the following steps: Obtain the historical mean value of the particle size correlation factor between adjacent regions; Analyze the change trend slope of the environmental parameter data; Take the product of the historical mean value and the change trend slope as the basic fluctuation amount; Normalize the basic fluctuation amount to obtain the particle distribution fluctuation index.

6. The method of claim 1, wherein, The method comprises the following steps: Retrieve the matching reference range of the particle distribution fluctuation index from the preset powder property database; Compare the difference between the particle distribution fluctuation index and the matching reference range; Calculate the initial abnormal value in each physical property dimension according to the size of the difference; Weighted average the initial abnormal value to obtain the abnormal probability of the metal powder in each physical property dimension.

7. The method of claim 6, wherein the metal powder property analysis method is used in an additive manufacturing process. The method comprises the following steps: Aggregate the abnormal probability values in all physical property dimensions; Calculate the variance and mean of the aggregated abnormal probability values; Based on the variance and mean, a comprehensive evaluation result of the metal powder characteristics is generated.

8. The method of claim 1, wherein, The adjustment of the additive manufacturing control parameters according to the comprehensive evaluation result of the metal powder characteristics comprises: Extracting key indicators in the comprehensive evaluation result of the metal powder characteristics; Comparing the key indicators with preset threshold values to generate control parameter adjustment amounts; Based on the control parameter adjustment amounts, updating the layer thickness and scanning speed parameters of the additive manufacturing equipment.

9. The method of claim 8, wherein the metal powder is a metal powder used in an additive manufacturing process. After the metal powder characteristic analysis is completed, further comprising: Monitoring the actual execution effect of the updated layer thickness and scanning speed parameters; Collecting the execution effect data as feedback input to the acquisition steps of the real-time image data and the environmental parameter data.

10. The method of claim 1, wherein, The method further comprises an iterative optimization step: After the comprehensive evaluation result of the metal powder characteristics is generated, the convergence state of the evaluation result is checked; When the convergence state does not meet the preset conditions, the calculation step of the particle distribution fluctuation indicator and the subsequent steps are repeated until the convergence state meets the preset conditions.

Images