Nano-metal powder particle size identification method and device

Abstract

The present application relates to the field of particle size identification technology, and discloses a particle size identification method and device for nano metal powder, wherein the method includes: obtaining network identification parameters matching the suspension medium; when generating the network identification parameters, constructing a standard suspension sample based on the suspension medium, and adding multiple different medium redundancies to the suspension medium to construct multiple groups of different comparative suspension samples; determining the environmental characteristics of the suspension sample, and analyzing the environmental characteristics to generate an environmental label corresponding to the suspension sample; using the environmental label as an auxiliary parameter and the network identification parameter as a main parameter to construct a particle size identification model suitable for the suspension sample; identifying the particle size of the nano metal powder in the suspension sample through the particle size identification model. The technical solution provided by the present application can improve the identification accuracy of the particle size of the nano metal powder.

Description

Technical Field

[0001] The present application relates to the technical field of particle size identification, and in particular to a method and device for identifying the particle size of nano metal powder. Background Art

[0002] Dynamic light scattering (DLS) is a high-precision technology that analyzes the fluctuations of laser scattered light from particles to deduce particle size information. It is widely used to measure the particle size and distribution of materials.

[0003] However, DLS measurements can be affected by medium properties (e.g., refractive index, viscosity) and environmental factors (e.g., temperature), which can change the characteristics of the scattered light and thus affect the accuracy of the particle size. Summary of the invention

[0004] The present application provides a method and device for identifying the particle size of nano metal powder, which achieves the technical effect of improving the identification accuracy of the particle size of nano metal powder.

[0005] In order to achieve the above objectives, the main technical solutions adopted in this application include:

[0006] In a first aspect, an embodiment of the present application provides a method for identifying the particle size of nano metal powders, the method comprising:

[0007] Obtaining a suspension sample of nano-metal powder and identifying a suspension medium of the suspension sample;

[0008] Acquire network recognition parameters that match the suspension medium; when generating the network recognition parameters, construct a standard suspension sample based on the suspension medium, and add a plurality of different medium redundancies to the suspension medium to construct a plurality of different comparative suspension samples; wherein when the network recognition parameters trained based on the standard suspension sample are applied to the comparative suspension sample, the obtained particle size prediction result is different from the standard particle size prediction result;

[0009] Determine the current environmental characteristics of the suspension sample, analyze the environmental characteristics, and generate an environmental label corresponding to the suspension sample;

[0010] Using the environmental label as an auxiliary parameter and the network identification parameter as a main parameter, a particle size identification model suitable for the suspension sample is constructed;

[0011] The particle size of the nano metal powder in the suspension sample is identified by the particle size identification model.

[0012] The present embodiment provides a nano-metal powder particle size identification method, which considers the influence of environmental conditions on particle size by combining network identification parameters with environmental labels, thereby improving the accuracy and reliability of particle size prediction. Through diversified training samples, especially by adding medium redundancy, the adaptability of the model to different media and environmental conditions is enhanced, ensuring the stability of particle size measurement in various changing environments.

[0013] In one embodiment, adding a plurality of different medium redundancies to the suspension medium to construct a plurality of different groups of comparative suspension samples comprises:

[0014] Determining a standard medium coefficient of the suspension medium, and adding medium redundancy on the basis of the standard medium coefficient to generate a plurality of different interfering medium coefficients;

[0015] The interfering suspension medium corresponding to each interfering medium coefficient is queried, and for any interfering suspension medium, an equal amount of nano-metal powder in the suspension sample is obtained, and a comparative suspension sample is constructed based on the equal amount of nano-metal powder and the interfering suspension medium.

[0016] This embodiment can simulate the changes of suspension under different environmental conditions by determining the standard medium coefficient of the suspension medium and adding redundancy to generate multiple interfering medium coefficients on this basis, thereby improving the robustness and adaptability of the particle size prediction model. Specifically, the standard medium coefficient provides a benchmark for the experiment to ensure the consistency of measurements under different conditions. By querying the interfering suspension medium corresponding to each interfering medium coefficient and obtaining an equal amount of nano-metal powder on this basis, multiple comparative suspension samples are constructed, effectively eliminating the error caused by the difference in sample size.

[0017] In one embodiment, the network recognition parameters obtained by training the standard suspension sample are generated in the following manner:

[0018] constructing a target sample set based on the standard suspension sample, and constructing a calibration sample set based on the comparative suspension sample, wherein the calibration sample set includes a plurality of comparative suspension samples with different medium redundancy amounts;

[0019] Using the target sample set to train a preset recognition model to obtain initial network parameters of the preset recognition model, wherein under the initial network parameters, the target sample set has a first particle size recognition result;

[0020] Processing the comparative suspension sample in the calibration sample set using a preset recognition model having the initial network parameters to obtain a second particle size recognition result;

[0021] Clustering the second particle size recognition result, and determining whether the first particle size recognition result is located in the cluster obtained by clustering, and if so, adjusting the initial network parameters so that after processing the target sample set based on the adjusted network parameters, the first particle size recognition result is still obtained, and after clustering the third particle size recognition results obtained after processing the correction sample set based on the adjusted network parameters, the first particle size recognition result is located outside the cluster;

[0022] The network parameters corresponding to the first particle size recognition result being outside the cluster are determined as network recognition parameters obtained through training based on the standard suspension sample.

[0023] This embodiment can significantly improve the accuracy and adaptability of the model, especially in applications under various media conditions, by analyzing the particle size recognition results of the target sample set and the calibration sample set and adjusting the initial network parameters. First, by processing the calibration sample set, the second particle size recognition result is obtained and cluster analysis is performed, which can effectively enhance the robustness of the model. If the first particle size recognition result is located in the cluster cluster, it means that the initial model fails to accurately distinguish the particle size recognition results under different media conditions, and thus it is necessary to further adjust the network parameters to ensure that the model can accurately identify the particle size in the target sample set. The adjusted network parameters enable the model to obtain clearly distinguished recognition results when processing standard suspension samples and calibration sample sets, thereby improving the classification accuracy and stability of the model. In addition, the optimized network parameters can also ensure the adaptability of the model under complex media conditions, ultimately ensuring that the recognition results of standard samples and calibration samples can be effectively separated and improving the generalization ability of the model.

[0024] In one embodiment, determining the current environmental characteristics of the suspension sample includes:

[0025] The temperature and humidity characteristics are collected according to a preset sampling period, and the collected temperature and humidity characteristics are arranged into a temperature sequence and a humidity sequence according to the collected timestamp;

[0026] The temperature features of the temperature sequence and the humidity features of the humidity sequence are extracted respectively, and the temperature features and the humidity features are concatenated into the environmental features of the current suspension sample.

[0027] This embodiment ensures that the temperature and humidity data have high time accuracy and consistency by accurately collecting temperature and humidity features and collating data according to the collected timestamps, providing reliable basic data for subsequent analysis. On this basis, the features of the temperature and humidity sequences are extracted separately and spliced ​​into a complete environmental feature, which can fully characterize the environmental state of the suspension sample. This feature integration improves the model's sensitivity to environmental changes, enhances its adaptability and prediction accuracy, and makes the model more accurate and robust under different environmental conditions.

[0028] In one embodiment, analyzing the environmental characteristics to generate an environmental label corresponding to the suspension sample includes:

[0029] Inputting the environmental features into the trained environment recognition model, so as to output the environment category value corresponding to the environmental features through the environment recognition model;

[0030] The category interval in which the environmental category value is located is determined, and a label in the category interval is used as an environmental label corresponding to the suspension sample, wherein the label value of the environmental label is greater than 0 and less than 1.

[0031] This embodiment can reflect the accurate environmental status of the sample by converting the environmental category value into an environmental label. At the same time, the adaptability to environmental changes is enhanced through flexible interval setting and label allocation. In addition, the label value is limited to the range of (0,1), which ensures standardization and consistency and avoids distortion or deviation. This method not only improves the accuracy and real-time response capability of environmental monitoring, but also reduces manual intervention and improves the efficiency of environmental control and monitoring. Accurate and automated determination of the environmental status of suspension samples is achieved.

[0032] In one embodiment, using the environmental label as an auxiliary parameter and the network identification parameter as a main parameter to construct a particle size identification model suitable for the suspension sample includes:

[0033] Determining a training strategy for a recognition model according to the label value of the environment label, the training strategy including at least one of a training momentum, a regularization parameter, a batch size, a learning rate decay, and an initialization method, and determining a training parameter for the recognition model according to the network recognition parameter, the training parameter including at least one of a weight, a bias, and a learning rate;

[0034] The recognition model that has completed the training strategy and training parameter configuration is determined as a particle size recognition model suitable for the suspension sample.

[0035] The present embodiment dynamically adjusts the training strategy and parameters of the recognition model by combining the environmental tag information, thereby significantly improving the training efficiency and accuracy of the model. According to the value of the environmental tag, the training strategies such as training momentum, regularization parameter, batch size, learning rate decay and initialization method are determined, and the key parameters such as the weight, bias and learning rate of the model are optimized by the network recognition parameter. Such adjustment ensures that the recognition model can automatically select the most suitable training method according to the specific environmental characteristics and particle size information of the suspension sample, thereby improving the adaptability and accuracy of the model, and avoiding the inefficiency or overfitting problem that the traditional static training method may bring. Further ensure that the trained model can be accurately applied to the particle size identification task of the suspension sample, improve the adaptive ability and intelligent level of the model, and realize a more efficient automatic particle size identification process. On the whole, this method not only optimizes the training process of the model, but also ensures the stability and high precision of the training results in practical applications.

[0036] In a second aspect, an embodiment of the present application provides a device for identifying the particle size of nano metal powders, the device comprising:

[0037] A sample acquisition unit, used to acquire a suspension sample of nano metal powder and identify a suspension medium of the suspension sample;

[0038] A parameter acquisition unit is used to acquire network recognition parameters that match the suspension medium; when the network recognition parameters are generated, a standard suspension sample is constructed based on the suspension medium, and a plurality of different medium redundancies are added to the suspension medium to construct a plurality of different comparative suspension samples; wherein when the network recognition parameters trained based on the standard suspension sample are applied to the comparative suspension sample, the obtained particle size prediction result is different from the standard particle size prediction result;

[0039] A label generating unit, used for determining the environmental characteristics of the suspension sample, analyzing the environmental characteristics, and generating an environmental label corresponding to the suspension sample;

[0040] A model building unit, used to use the environmental label as an auxiliary parameter and the network identification parameter as a main parameter to build a particle size identification model suitable for the suspension sample;

[0041] The particle size recognition unit is used to recognize the particle size of the nano metal powder in the suspension sample through the particle size recognition model.

[0042] In one embodiment, the parameter acquisition unit also includes a comparison sample construction module, which is used to determine the standard medium coefficient of the suspension medium and add medium redundancy on the basis of the standard medium coefficient to generate multiple different interference medium coefficients; query the interfering suspension medium corresponding to each of the interfering medium coefficients, and for any interfering suspension medium, obtain an equal amount of nano-metal powder in the suspension sample, and construct a comparison suspension sample based on the equal amount of nano-metal powder and the interfering suspension medium.

[0043] In one embodiment, the parameter acquisition unit is specifically used to construct a target sample set based on the standard suspension sample, and to construct a calibration sample set based on the comparative suspension sample, wherein the calibration sample set includes a plurality of comparative suspension samples with different medium redundancies; to train a preset recognition model using the target sample set to obtain initial network parameters of the preset recognition model, wherein under the initial network parameters, the target sample set has a first particle size recognition result; and to process the comparative suspension samples in the calibration sample set using the preset recognition model having the initial network parameters to obtain a second particle size recognition result. The method comprises the following steps: clustering the second particle size recognition result and determining whether the first particle size recognition result is located in the cluster cluster obtained by clustering. If so, adjusting the initial network parameters so that after processing the target sample set based on the adjusted network parameters, the first particle size recognition result is still obtained, and after clustering the third particle size recognition result obtained after processing the correction sample set based on the adjusted network parameters, the first particle size recognition result is located outside the cluster cluster; determining the network parameters corresponding to the first particle size recognition result when it is located outside the cluster cluster as the network recognition parameters obtained based on the training of the standard suspension sample.

[0044] In one embodiment, the label generation unit is specifically used to collect temperature and humidity characteristics according to a preset sampling period, and arrange the collected temperature and humidity characteristics into a temperature sequence and a humidity sequence according to the collected timestamp; respectively extract the temperature characteristics of the temperature sequence and the humidity characteristics of the humidity sequence, and splice the temperature characteristics and the humidity characteristics into the environmental characteristics of the current suspension sample. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the specific implementation methods of the present application or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0046] Figure 1 A flow chart of a method for identifying the particle size of nano metal powder provided in an embodiment of the present application;

[0047] Figure 2 A flowchart for constructing multiple groups of different comparative suspension samples provided in the embodiments of the present application;

[0048] Figure 3 A flowchart of a method for generating network identification parameters provided in an embodiment of the present application;

[0049] Figure 4 A flow chart for determining environmental characteristics provided in an embodiment of the present application;

[0050] Figure 5 A flowchart of generating environment labels provided in an embodiment of the present application;

[0051] Figure 6 A flowchart of step S7 provided in an embodiment of the present application;

[0052] Figure 7 A block diagram of a nano-metal powder particle size identification device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0053] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present application.

[0054] Dynamic Light Scattering (DLS) is a high-precision analytical technique that is widely used in materials science, biology, chemistry and other fields, especially in the measurement of particle size and distribution. Its basic principle is based on the scattering of light by particles after laser irradiation of the sample. By analyzing the intensity fluctuations of the scattered light and combining the Brownian motion of the particles in the liquid, the size information of the particles is derived. In this way, DLS can provide detailed information about the particle size distribution, especially for the measurement of nano-scale particles.

[0055] However, in practical applications, DLS technology also faces some challenges, one of the most important influencing factors is the physical properties of the medium. The properties of the medium solution, such as the refractive index and viscosity, directly affect the intensity and pattern of light scattering, thereby affecting the accurate measurement of particle size. Since different media have different scattering characteristics, when using DLS, corresponding calibration is required for different media to ensure the accuracy of the measurement results.

[0056] In addition, environmental factors, especially temperature changes, also have a significant impact on DLS measurements. Temperature not only affects the diffusion rate of particles in liquids, but also changes the viscosity of the solvent. These changes will cause changes in the fluctuation characteristics of scattered light, which in turn affects the calculation of particle size. Therefore, temperature fluctuations need to be strictly controlled to ensure the accuracy and consistency of measurements.

[0057] In view of these influencing factors, how to improve the accuracy of nanometal powder particle size identification has become the focus of current research.

[0058] In order to solve the above technical problems, according to an embodiment of the present application, an embodiment of a method for identifying the particle size of nano metal powder is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0059] In this embodiment, a method for identifying the particle size of nano metal powder is provided. Figure 1 A flow chart of a method for identifying the particle size of nano metal powder provided in an embodiment of the present application, such as Figure 1 As shown, the process includes the following steps:

[0060] Step S1, obtaining a suspension sample of nano metal powder and identifying a suspension medium of the suspension sample.

[0061] Nano metal powder refers to metal powder with a particle size of 1-100 nanometers at the nanometer level. These nanoparticles have unique physical and chemical properties due to their extremely small size, such as high specific surface area and high reactivity. Suspension samples refer to mixtures formed by uniform dispersion of nano metal powder in liquid medium. Suspension medium refers to the liquid used to disperse nano metal powder, which can be organic solvents, inorganic solvents or other liquids.

[0062] Specifically, it is necessary to obtain a suspension sample of nanometal powder, which is usually prepared in a laboratory by mixing the nanometal powder with an appropriate dispersion medium. For example, oil, alcohol, etc. can be used as the dispersion medium. Since different media will affect the dispersion state and particle size measurement results of the nanometal powder, chemical analysis can be used to identify the suspension medium in the suspension sample using infrared spectroscopy (FTIR), nuclear magnetic resonance (NMR) and other technical methods.

[0063] Step S3, obtaining network recognition parameters that match the suspension medium; when generating the network recognition parameters, a standard suspension sample is constructed based on the suspension medium, and a plurality of different medium redundancies are added to the suspension medium to construct a plurality of different comparative suspension samples; wherein, when the network recognition parameters trained according to the standard suspension sample are applied to the comparative suspension sample, the obtained particle size prediction result is different from the standard particle size prediction result.

[0064] Network identification parameters refer to key values ​​in a machine learning model that are used to define the structure and behavior of the model. Network identification parameters include weights, biases, activation function parameters, etc., which are optimized through the training process so that the model can accurately identify patterns that match the suspension medium. Standard suspension samples are suspension samples prepared under specific conditions, in which the particle size and distribution of the nanometal powder and the suspension medium are known, and are used to train machine learning models so that the model can learn how to accurately predict the particle size of the nanometal powder in the suspension medium. Comparative suspension samples are prepared on the basis of standard suspension samples by adding different amounts of medium redundancy, and are used to test and calibrate the model to ensure that the network identification parameters can accurately predict the particle size when applied to the model for particle size identification.

[0065] Specifically, standard suspension samples are prepared based on the suspension medium. The particle size of the nanometal powder in these samples is known and can be used as the basis for training the model. The machine learning model is trained using standard suspension samples. The model optimizes its network recognition parameters by learning the scattered light characteristics of the nanometal powder dissolved in the suspension medium in the standard suspension samples and the known particle size information to improve the accuracy of the prediction. Different amounts of medium redundancy (such as salt, organic solvent or inorganic solvent, etc.) are added to the suspension medium. These substances will change the physical and chemical properties of the suspension medium, thereby affecting the behavior of the nanometal powder in the suspension medium. For example, medium redundancy may change the density, viscosity and surface tension of the solution, all of which will affect the dispersion state and Brownian motion of the particles. Therefore, the scattered light characteristics of the nanometal powder in the suspension medium will also be affected, resulting in errors in particle size prediction. By introducing interference during the training process, that is, comparing suspension samples, the network recognition parameters can be made more robust, that is, the model can better identify and predict the particle size when facing standard suspension samples.

[0066] Step S5, determining the current environmental characteristics of the suspension sample, analyzing the environmental characteristics, and generating an environmental label corresponding to the suspension sample.

[0067] Environmental characteristics refer to various physical and chemical properties of the environment in which the suspension sample is located, such as temperature, humidity, pH value, light intensity, etc. These characteristics have an important influence on the behavior and properties of particles in the suspension.

[0068] Specifically, various sensor devices such as temperature and humidity sensors, pH meters, etc. are used to collect relevant data of the environment in which the suspension sample is located in real time, and the collected data is recorded and preliminarily analyzed to determine the characteristic values ​​of the current environment, such as temperature range, humidity level, etc. Based on the analysis results, a corresponding environmental label is generated. This label can be a numerical range, a classification label, or a comprehensive feature vector to describe the specific environmental conditions of the suspension sample.

[0069] Step S7, using the environmental label as an auxiliary parameter and the network recognition parameter as a main parameter to construct a particle size recognition model suitable for the suspension sample.

[0070] Specifically, the environmental label is used as an auxiliary parameter and combined with the network identification parameter. The environmental label provides detailed information about the environment in which the suspension sample is located, while the network identification parameter contains the model's ability to predict the particle size of nanometal powder under the conditions of the suspension medium. The model is trained using a data set containing environmental labels and network identification parameters. During the training process, the model will learn how to accurately predict the particle size of nanometal powder under different environmental conditions, and optimize its prediction performance under different environmental conditions by adjusting the structure and parameters of the model. For example, methods such as cross-validation can be used to evaluate the generalization ability of the model, and parameters can be adjusted based on the evaluation results.

[0071] Step S9, identifying the particle size of the nano metal powder in the suspension sample by using a particle size identification model.

[0072] Specifically, for a suspension sample prepared with any suspension medium, a trained particle size recognition model matching the suspension medium is used to identify the particle size of the nano metal powder.

[0073] The present embodiment provides a nano-metal powder particle size identification method, which considers the influence of environmental conditions on particle size by combining network identification parameters with environmental labels, thereby improving the accuracy and reliability of particle size prediction. Through diversified training samples, especially by adding medium redundancy, the adaptability of the model to different media and environmental conditions is enhanced, ensuring the stability of particle size measurement in various changing environments.

[0074] Figure 2 A flowchart of constructing multiple groups of different comparison suspension samples provided in an embodiment of the present application may include the following steps:

[0075] Step S31, determining a standard medium coefficient of the suspension medium, and adding medium redundancy on the basis of the standard medium coefficient to generate a plurality of different interfering medium coefficients.

[0076] Specifically, the standard medium coefficient is a key parameter that describes the physical properties of the suspension medium, including properties such as density and viscosity. These physical properties directly affect the movement state of the nanometal powder in the suspension medium, especially the characteristics of its Brownian motion. For example, when the density of the suspension medium is high, the movement speed of the nanometal powder will slow down, thereby affecting its scattering properties. The high viscosity of the suspension medium will increase the interaction force between the nanometal powders, thereby affecting their dispersibility and may change the intensity of the scattered light. The Brownian motion of the nanometal powder in the solution causes its particles to constantly collide and move, and this dynamic change causes fluctuations in the light scattering intensity. The scattered light intensity is closely related to the size and concentration of the nanoparticles and the physical properties of the medium. Therefore, the standard medium coefficient plays a vital role in the light scattering measurement of the suspension.

[0077] Medium redundancy refers to the additional types or amounts of suspending media added to the suspension medium. These additional substances can be salts, organic solvents, surfactants, etc., which will change the density, viscosity, surface tension and other physical properties of the original suspension sample suspension medium, thereby affecting the overall behavior of the suspension. Changes in the types and amounts of redundant substances will cause changes in the light scattering characteristics of the suspension and produce different interfering medium coefficients.

[0078] The interfering medium coefficient reflects how the light scattering properties of the suspension change after the addition of redundant substances. By adjusting the amount of redundancy and the type of redundant substances, a series of media with different physical properties and scattering properties can be generated. Each interfering medium coefficient represents the effect of a specific medium formulation on the dispersion, Brownian motion and light scattering of nanometal powders. By adding different types of redundant substances (such as salts, surfactants or organic solvents), the physical properties of the suspension can be adjusted and different interfering medium coefficients can be generated. These changes lead to changes in particle dispersion, Brownian motion and scattered light intensity, which affects the prediction and measurement of particle size.

[0079] Step S33, querying the interfering suspension medium corresponding to each interfering medium coefficient, obtaining an equal amount of nano-metal powder in the suspension sample for any interfering suspension medium, and constructing a comparative suspension sample based on the equal amount of nano-metal powder and the interfering suspension medium.

[0080] Specifically, determine the specific values ​​of each interfering medium coefficient. These coefficients are obtained by adding different amounts of medium redundancy to the suspension medium and then re-measuring, reflecting the changes in the physical properties of the medium under different conditions. According to the interfering medium coefficient, query the specific composition and properties of the corresponding interfering suspension medium. For example, if the interfering medium coefficient indicates that the viscosity has increased by a certain multiple, then the corresponding interfering suspension medium may be a polymer solution with a specific concentration added. Accurately weigh equal amounts of nanometal powder from the suspension sample to ensure that the amount of nanometal powder remains consistent under different medium conditions. Mix equal amounts of nanometal powder with the interfering suspension medium queried, stir thoroughly to make it evenly dispersed, and thus construct a comparative suspension sample. For example, nanometal powder can be mixed with a polymer solution added to form a new suspension sample. Repeat the above operation for different interfering medium coefficients to construct multiple groups of different comparative suspension samples for subsequent particle size identification and analysis.

[0081] This embodiment can simulate the changes of suspension under different environmental conditions by determining the standard medium coefficient of the suspension medium and adding redundancy to generate multiple interfering medium coefficients on this basis, thereby improving the robustness and adaptability of the particle size prediction model. Specifically, the standard medium coefficient provides a benchmark for the experiment to ensure the consistency of measurements under different conditions. By querying the interfering suspension medium corresponding to each interfering medium coefficient and obtaining an equal amount of nano-metal powder on this basis, multiple comparative suspension samples are constructed, effectively eliminating the error caused by the difference in sample size.

[0082] Figure 3 A flowchart of a method for generating network identification parameters provided in an embodiment of the present application, the process may include the following steps:

[0083] Step S331 : constructing a target sample set based on the standard suspension sample, and constructing a calibration sample set based on the comparative suspension sample, wherein the calibration sample set includes a plurality of comparative suspension samples with different medium redundancies.

[0084] Specifically, a target sample set is constructed based on standard suspension samples, in which the particle size of the nanometal powder is known and the properties of the suspension medium match the standard medium coefficient. The target sample set is used to train the model so that it can accurately predict the particle size under standard medium conditions. A calibration sample set is constructed based on comparative suspension samples, which includes a variety of comparative suspension samples with different medium redundancies. The calibration sample set is used to test and calibrate the model to ensure its prediction accuracy under different medium conditions.

[0085] Step S333, using the target sample set to train the preset recognition model to obtain initial network parameters of the preset recognition model, wherein under the initial network parameters, the target sample set has a first particle size recognition result.

[0086] Specifically, a suitable machine learning or deep learning model is selected as the preset recognition model, such as a support vector machine, a neural network, etc. The input of the model is the characteristic data of the suspension sample in the target sample set, such as the Brownian motion of the nanometal powder in the solution, and the sequence data that the intensity of the scattered light changes over time, and the output is the particle size of the nanometal powder. The preset recognition model is trained using the target sample set, and the network parameters of the model are adjusted through the optimization algorithm so that it can accurately predict the particle size in the target sample set. After the training is completed, the initial network parameters of the preset recognition model are obtained. Under the initial network parameters, the target sample set is processed to obtain the first particle size recognition result. This result is the recognition result of the model for samples of known particle size under standard medium conditions.

[0087] Step S335 , using a preset recognition model with initial network parameters to process the comparative suspension sample in the calibration sample set to obtain a second particle size recognition result.

[0088] Specifically, the comparison suspension samples in the calibration sample set are processed using a preset recognition model with initial network parameters, that is, the characteristic data of the calibration sample set is input, and the model outputs the corresponding particle size recognition result to obtain the second particle size recognition result. This result reflects the particle size recognition of the model under different medium redundancy conditions, which may be different from the first particle size recognition result of the target sample set because the medium properties of the calibration sample set are different from those of the target sample set.

[0089] Step S337, clustering the second particle size recognition result, and determining whether the first particle size recognition result is located in the cluster cluster obtained by clustering. If so, adjusting the initial network parameters so that after processing the target sample set based on the adjusted network parameters, the first particle size recognition result is still obtained, and after clustering the third particle size recognition result obtained after processing the correction sample set based on the adjusted network parameters, the first particle size recognition result is located outside the cluster cluster.

[0090] Specifically, cluster analysis is performed on the second particle size recognition result, and similar recognition results are grouped into the same cluster. The clustering algorithm can be K-means, hierarchical clustering, etc. The purpose is to discover the distribution pattern of particle size recognition results under different medium conditions. Determine whether the first particle size recognition result is located in the cluster obtained by clustering. If it is located in the cluster, it means that there is a certain similarity between the model's recognition results of the particle size under different medium conditions and the results under standard medium conditions, which means that the model may be affected by different medium conditions and produce similar prediction results. At this time, the recognition results under standard medium conditions may not be accurate enough, and the model needs to be further optimized.

[0091] The network parameters are adjusted according to the clustering results to avoid interference of different medium conditions on the model and improve the adaptability and robustness of the model. The goal of the adjustment is to obtain the first particle size recognition result after processing the target sample set based on the adjusted network parameters, that is, to maintain the recognition accuracy under standard medium conditions; at the same time, after processing the correction sample set based on the adjusted network parameters, the third particle size recognition result is obtained after clustering, and the first particle size recognition result is located outside the cluster cluster, that is, the recognition results under different medium conditions can be clearly distinguished, improving the robustness and adaptability of the model. The network parameters can be adjusted by increasing the number of network layers, improving the activation function, adjusting the learning rate, etc.

[0092] Step S339, determining the network parameters corresponding to the first particle size recognition result being outside the cluster as the network recognition parameters obtained by training based on the standard suspension sample.

[0093] Specifically, when the first particle size recognition result is outside the cluster, it means that the adjusted network parameters enable the model to effectively distinguish the particle size recognition results under different medium conditions, and still maintain a high recognition accuracy under standard medium conditions. The corresponding network parameters at this time are determined as the network recognition parameters obtained by training based on the standard suspension sample, which will be used in the subsequent nano-metal powder particle size recognition task to improve the accuracy and reliability of recognition.

[0094] This embodiment can significantly improve the accuracy and adaptability of the model, especially in applications under various media conditions, by analyzing the particle size recognition results of the target sample set and the calibration sample set and adjusting the initial network parameters. First, by processing the calibration sample set, the second particle size recognition result is obtained and cluster analysis is performed, which can effectively enhance the robustness of the model. If the first particle size recognition result is located in the cluster cluster, it means that the initial model fails to accurately distinguish the particle size recognition results under different media conditions, and thus it is necessary to further adjust the network parameters to ensure that the model can accurately identify the particle size in the target sample set. The adjusted network parameters enable the model to obtain clearly distinguished recognition results when processing standard suspension samples and calibration sample sets, thereby improving the classification accuracy and stability of the model. In addition, the optimized network parameters can also ensure the adaptability of the model under complex media conditions, ultimately ensuring that the recognition results of standard samples and calibration samples can be effectively separated and improving the generalization ability of the model.

[0095] Figure 4 A flowchart for determining environmental characteristics provided in an embodiment of the present application may include the following steps:

[0096] Step S51, collecting temperature and humidity characteristics according to a preset sampling period, and arranging the collected temperature and humidity characteristics into a temperature sequence and a humidity sequence according to the collected timestamps.

[0097] Specifically, the sampling period can be set to collect temperature and humidity data once every minute or every 5 minutes. Of course, it can also be determined according to the experimental requirements and environmental changes to ensure that subtle changes in ambient temperature and humidity can be captured. Use temperature and humidity sensors and other equipment to collect temperature and humidity data of the environment in which the suspension sample is located in real time according to the preset sampling period. The sensor should be placed near the sample to ensure that the collected data can accurately reflect the environmental conditions of the sample. When collecting each temperature and humidity data point, record the corresponding timestamp, that is, the specific time of data collection. The timestamp can be a combination of date and time for subsequent data analysis and processing. According to the collected timestamps, the collected temperature data is arranged in chronological order to form a temperature sequence; similarly, the humidity data is arranged in chronological order to form a humidity sequence. These two sequences can be represented as one-dimensional arrays or lists, in which each element corresponds to a temperature or humidity value at a time point.

[0098] Step S53 , extracting the temperature features of the temperature sequence and the humidity features of the humidity sequence respectively, and combining the temperature features and the humidity features into the environmental features of the current suspension sample.

[0099] Specifically, feature extraction is performed on the temperature sequence, for example, including the average temperature, maximum temperature, minimum temperature, standard deviation of temperature, etc. These features can reflect the overall level and fluctuation of the temperature sequence. Feature extraction is performed on the humidity sequence, for example, including the average humidity, maximum humidity, minimum humidity, standard deviation of humidity, etc. The extracted temperature features and humidity features are spliced ​​to form a comprehensive environmental feature vector. The splicing method can be a simple splicing, that is, the values ​​of the temperature features and humidity features are arranged in sequence to form a longer feature vector. This environmental feature vector contains a variety of feature information of temperature and humidity, and can comprehensively describe the environmental conditions in which the suspension sample is currently located.

[0100] This embodiment ensures that the temperature and humidity data have high time accuracy and consistency by accurately collecting temperature and humidity features and collating data according to the collected timestamps, providing reliable basic data for subsequent analysis. On this basis, the features of the temperature and humidity sequences are extracted separately and spliced ​​into a complete environmental feature, which can fully characterize the environmental state of the suspension sample. This feature integration improves the model's sensitivity to environmental changes, enhances its adaptability and prediction accuracy, and makes the model more accurate and robust under different environmental conditions.

[0101] Figure 5 A flowchart of generating an environment tag provided in an embodiment of the present application may include the following steps:

[0102] Step S531, inputting the environmental features into the trained environment recognition model, so as to output the environment category value corresponding to the environmental features through the environment recognition model.

[0103] Specifically, the environmental recognition model is a trained machine learning model, whose main function is to identify and classify the category of the environment based on the input environmental features. The training data of the model should contain feature data under a variety of different environmental conditions and their corresponding category labels, so that the model can learn the mapping relationship between different environmental features and categories. The environmental feature vector is used as input and input into the environmental recognition model. This environmental feature vector contains a variety of feature information of temperature and humidity, which can comprehensively describe the current environmental conditions of the suspension sample. After processing by the environmental recognition model, an environmental category value will be output. This value is a numerical value that represents the model's recognition result of the current environmental features and reflects the category or state of the environment in which the sample is located.

[0104] Step S533 , determining the category interval where the environmental category value is located, and using the label of the category interval as the environmental label corresponding to the suspension sample, wherein the label value of the environmental label is greater than 0 and less than 1.

[0105] Specifically, the category interval refers to dividing the possible value range of the environmental category value into several continuous intervals, each of which corresponds to a specific environmental category or state. For example, the range of environmental category values ​​can be divided into intervals such as [0, 0.2), [0.2, 0.4), [0.4, 0.6), [0.6, 0.8), [0.8, 1.0], and each interval corresponds to a different environmental label. The label value of the environmental label is greater than 0 and less than 1, that is, the label value is in the open interval of (0, 1). The label value within this range can represent different degrees or states of the environment, such as high humidity, moderate temperature, etc. According to the environmental category value output by the environmental recognition model, the category interval in which it is located is determined. For example, if the output environmental category value is 0.35, it is in the category interval of [0.2, 0.4). The label of the determined category interval is used as the environmental label corresponding to the suspension sample. This environmental label can accurately reflect the characteristics and state of the environment in which the sample is located. For example, the interval [0.2, 0.4) may correspond to an environmental label with a label value of 0.3, indicating that the humidity of the environment in which the sample is located is high.

[0106] This embodiment can reflect the accurate environmental status of the sample by converting the environmental category value into an environmental label. At the same time, the adaptability to environmental changes is enhanced through flexible interval setting and label allocation. In addition, the label value is limited to the range of (0,1), which ensures standardization and consistency and avoids distortion or deviation. This method not only improves the accuracy and real-time response capability of environmental monitoring, but also reduces manual intervention and improves the efficiency of environmental control and monitoring. Accurate and automated determination of the environmental status of suspension samples is achieved.

[0107] Figure 6 The flowchart of step S7 provided in the embodiment of the present application may include the following steps:

[0108] Step S71, determining the training strategy of the recognition model according to the label value of the environment label, the training strategy includes at least one of training momentum, regularization parameter, batch size, learning rate decay and initialization method, and determining the recognition model training parameters according to the network recognition parameters, the training parameters include at least one of weight, bias and learning rate.

[0109] Specifically, the label value of the environment label can be used to adjust the size of the training momentum. A larger label value may mean that the environment in which the sample is located is more complex. At this time, the training momentum can be increased to help the model converge faster during the gradient descent process and help escape the local minimum.

[0110] The environmental label value can also be used to adjust the size of the regularization parameter. A larger label value may indicate that the environment is noisy. At this time, the regularization parameters (such as L1 and L2 regularization coefficients) can be increased to prevent the model from overfitting and improve its generalization ability in complex environments.

[0111] The batch size can be adjusted dynamically based on the environment label value. A smaller label value may indicate a simpler environment, and a smaller batch size can be used for training so that the model can learn sample features more carefully. When the label value is large, the batch size can be appropriately increased to improve training efficiency.

[0112] The environment label value can be used to formulate a learning rate decay strategy. For samples with larger label values, a faster learning rate decay rate can be set to avoid large fluctuations in the model in complex environments; and for samples with smaller label values, a slower learning rate decay rate can be set so that the model can learn the sample features more fully.

[0113] Choose different parameter initialization methods based on the environment label value. For example, when the label value is small, you can use a small random number to initialize the weight and bias to avoid the gradient vanishing problem caused by the initial parameter being too large; when the label value is large, you can use methods such as Xavier initialization to help the model converge better in complex environments.

[0114] The network recognition parameters include weights trained under standard suspension sample conditions, which can be used as the initial weights of the recognition model. The initial weights are updated through the back-propagation algorithm to minimize the loss function, thereby improving the model's accuracy in identifying particle size.

[0115] The bias in the network recognition parameters can also be used as the initial bias of the recognition model. The bias is updated through the back propagation algorithm to adjust the activation threshold of the neuron and help the model better fit the data.

[0116] The learning rate can be determined based on the network identification parameters and the environmental label value. The learning rate in the network identification parameters is obtained based on standard sample training and can be used as the basic learning rate; the learning rate is then fine-tuned in combination with the environmental label value to adapt to the training requirements under different environmental conditions. Dynamically adjusting the learning rate or using a learning rate decay strategy can optimize the model training process.

[0117] Step S73: determining the recognition model that has completed the training strategy and training parameter configuration as a particle size recognition model suitable for the suspension sample.

[0118] Specifically, the training strategy determined by the environmental label and the training parameters determined by the network recognition parameters are configured into the recognition model to complete the model configuration process. The reasonable configuration of these strategies and parameters can make the model better adapt to the different environmental conditions of the suspension samples. After training and optimization, the configured recognition model can accurately identify the particle size of nanometal powder. This model will be used as the final particle size recognition model suitable for suspension samples in actual particle size recognition tasks to ensure that reliable recognition results can be obtained under different environmental conditions.

[0119] The present embodiment dynamically adjusts the training strategy and parameters of the recognition model by combining the environmental tag information, thereby significantly improving the training efficiency and accuracy of the model. According to the value of the environmental tag, the training strategies such as training momentum, regularization parameter, batch size, learning rate decay and initialization method are determined, and the key parameters such as the weight, bias and learning rate of the model are optimized by the network recognition parameter. Such adjustment ensures that the recognition model can automatically select the most suitable training method according to the specific environmental characteristics and particle size information of the suspension sample, thereby improving the adaptability and accuracy of the model, and avoiding the inefficiency or overfitting problem that the traditional static training method may bring. Further ensure that the trained model can be accurately applied to the particle size identification task of the suspension sample, improve the adaptive ability and intelligent level of the model, and realize a more efficient automatic particle size identification process. On the whole, this method not only optimizes the training process of the model, but also ensures the stability and high precision of the training results in practical applications.

[0120] Accordingly, please refer to Figure 7 A block diagram of a nano-metal powder particle size identification device provided in an embodiment of the present application, the device comprising:

[0121] The sample acquisition unit 101 is used to acquire a suspension sample of nano-metal powder and identify a suspension medium of the suspension sample;

[0122] The parameter acquisition unit 103 is used to acquire network recognition parameters that match the suspension medium; when the network recognition parameters are generated, a standard suspension sample is constructed based on the suspension medium, and a plurality of different medium redundancies are added to the suspension medium to construct a plurality of different comparative suspension samples; wherein when the network recognition parameters obtained by training the standard suspension sample are applied to the comparative suspension sample, the obtained particle size prediction result is different from the standard particle size prediction result;

[0123] The label generation unit 105 is used to determine the environmental characteristics of the suspension sample, analyze the environmental characteristics, and generate an environmental label corresponding to the suspension sample;

[0124] A model building unit 107, used to use the environmental label as an auxiliary parameter and the network identification parameter as a main parameter to build a particle size identification model suitable for the suspension sample;

[0125] The particle size recognition unit 109 is used to recognize the particle size of the nano metal powder in the suspension sample through a particle size recognition model.

[0126] In some optional embodiments, the parameter acquisition unit 103 also includes a comparison sample construction module, which is used to determine the standard medium coefficient of the suspension medium and add medium redundancy on the basis of the standard medium coefficient to generate multiple different interference medium coefficients; query the interfering suspension medium corresponding to each interfering medium coefficient, and for any interfering suspension medium, obtain an equal amount of nano-metal powder in the suspension sample, and construct a comparison suspension sample based on equal amounts of nano-metal powder and interfering suspension medium.

[0127] In some optional embodiments, the parameter acquisition unit 103 is specifically used to construct a target sample set based on the standard suspension sample, and to construct a correction sample set based on the comparative suspension sample, wherein the correction sample set includes a plurality of comparative suspension samples with different medium redundancy; train a preset recognition model using the target sample set to obtain initial network parameters of the preset recognition model, wherein under the initial network parameters, the target sample set has a first particle size recognition result; process the comparative suspension samples in the correction sample set using the preset recognition model with the initial network parameters to obtain a second particle size recognition result; cluster the second particle size recognition result, and determine whether the first particle size recognition result is located in the cluster cluster obtained by clustering, and if so, adjust the initial network parameters so that after processing the target sample set based on the adjusted network parameters, the first particle size recognition result is still obtained, and after clustering the third particle size recognition result obtained after processing the correction sample set based on the adjusted network parameters, the first particle size recognition result is located outside the cluster cluster; determine the network parameters corresponding to the first particle size recognition result being located outside the cluster cluster as the network recognition parameters obtained by training based on the standard suspension sample.

[0128] In some optional embodiments, the label generation unit 105 is specifically used to collect temperature and humidity characteristics according to a preset sampling period, and arrange the collected temperature and humidity characteristics into a temperature sequence and a humidity sequence according to the collected timestamp; respectively extract the temperature characteristics of the temperature sequence and the humidity characteristics of the humidity sequence, and splice the temperature characteristics and the humidity characteristics into the current environmental characteristics of the suspension sample.

[0129] In some optional embodiments, the label generation unit 105 is specifically used to input the environmental features into a trained environmental recognition model to output the environmental category value corresponding to the environmental features through the environmental recognition model; determine the category interval in which the environmental category value is located, and use the label possessed by the category interval as the environmental label corresponding to the suspension sample, wherein the label value of the environmental label is greater than 0 and less than 1.

[0130] In some optional embodiments, the model building unit 107 is specifically used to determine the training strategy of the recognition model according to the label value of the environmental label, the training strategy includes at least one of training momentum, regularization parameter, batch size, learning rate decay and initialization method, and determine the recognition model training parameters according to the network recognition parameters, the training parameters include at least one of weight, bias and learning rate; the recognition model that completes the training strategy and training parameter configuration is determined as a particle size recognition model suitable for suspension samples.

[0131] The further functional description of each of the above modules and units is the same as that of the above corresponding embodiments and will not be repeated here.

[0132] In this embodiment, a nano-metal powder particle size identification device is presented in the form of a functional unit, where the unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and memory that executes one or more software or fixed programs, and / or other devices that can provide the above functions.

[0133] Although the embodiments of the present application are described in conjunction with the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the present application, and such modifications and variations are all within the scope defined by the appended claims.

[0134] The methods or devices described in the above embodiments may be implemented by a computer chip or entity, or by a product having a certain function. A typical implementation device is a computer. Specifically, the computer may be, for example, a personal computer, a laptop computer, a cellular phone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.

[0135] For the convenience of description, the above device is described in various units according to their functions. Of course, when implementing the present application, the functions of each unit can be implemented in the same or multiple software and / or hardware.

[0136] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods or devices. Therefore, the present application may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0137] The present application is described with reference to the flowcharts and / or block diagrams of the methods and devices according to the embodiments of the present application. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of the processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0138] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0139] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0140] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0141] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0142] The above is only an embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims of the present application.

[0143] Although the embodiments of the present application have been described in conjunction with the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the present application, and such modifications and variations are all within the scope defined by the appended claims.

Claims

1. A method for identifying the particle size of nano metal powder, characterized in that: The method comprises: Obtaining a suspension sample of nano-metal powder and identifying a suspension medium of the suspension sample; Acquire network identification parameters that match the suspension medium; when generating the network identification parameters, construct a standard suspension sample based on the suspension medium, and add a plurality of different medium redundancies to the suspension medium to construct a plurality of different comparative suspension samples; wherein when the network identification parameters trained based on the standard suspension sample are applied to the comparative suspension sample, the obtained particle size prediction result is different from the standard particle size prediction result; the medium redundancy characterizes the additional different types of suspension media added to the suspension medium; Determine the current environmental characteristics of the suspension sample, analyze the environmental characteristics, and generate an environmental label corresponding to the suspension sample; Using the environmental label as an auxiliary parameter and the network identification parameter as a main parameter, a particle size identification model suitable for the suspension sample is constructed; The particle size of the nano metal powder in the suspension sample is identified by the particle size identification model.

2. The method according to claim 1, characterized in that Adding a plurality of different medium redundancies to the suspension medium to construct a plurality of different groups of comparative suspension samples comprises: Determining a standard medium coefficient of the suspension medium, and adding medium redundancy on the basis of the standard medium coefficient to generate a plurality of different interfering medium coefficients; The interfering suspension medium corresponding to each interfering medium coefficient is queried, and for any interfering suspension medium, an equal amount of nano-metal powder in the suspension sample is obtained, and a comparative suspension sample is constructed based on the equal amount of nano-metal powder and the interfering suspension medium.

3. The method according to claim 1 or 2, characterized in that: The network recognition parameters obtained by training the standard suspension sample are generated in the following manner: constructing a target sample set based on the standard suspension sample, and constructing a calibration sample set based on the comparative suspension sample, wherein the calibration sample set includes a plurality of comparative suspension samples with different medium redundancy amounts; Using the target sample set to train a preset recognition model to obtain initial network parameters of the preset recognition model, wherein under the initial network parameters, the target sample set has a first particle size recognition result; Processing the comparative suspension sample in the calibration sample set using a preset recognition model having the initial network parameters to obtain a second particle size recognition result; Clustering the second particle size recognition result, and determining whether the first particle size recognition result is located in the cluster obtained by clustering, and if so, adjusting the initial network parameters so that after processing the target sample set based on the adjusted network parameters, the first particle size recognition result is still obtained, and after clustering the third particle size recognition results obtained after processing the correction sample set based on the adjusted network parameters, the first particle size recognition result is located outside the cluster; The network parameters corresponding to the first particle size recognition result being outside the cluster are determined as network recognition parameters obtained through training based on the standard suspension sample.

4. The method according to claim 1, characterized in that Determining the current environmental characteristics of the suspension sample includes: The temperature and humidity characteristics are collected according to a preset sampling period, and the collected temperature and humidity characteristics are arranged into a temperature sequence and a humidity sequence according to the collected timestamp; The temperature features of the temperature sequence and the humidity features of the humidity sequence are extracted respectively, and the temperature features and the humidity features are concatenated into the environmental features of the current suspension sample.

5. The method according to claim 1 or 4, characterized in that: Analyzing the environmental characteristics to generate an environmental label corresponding to the suspension sample includes: Inputting the environmental features into the trained environment recognition model, so as to output the environment category value corresponding to the environmental features through the environment recognition model; The category interval in which the environmental category value is located is determined, and a label in the category interval is used as an environmental label corresponding to the suspension sample, wherein the label value of the environmental label is greater than 0 and less than 1.

6. The method according to claim 1, characterized in that Using the environmental label as an auxiliary parameter and the network identification parameter as a main parameter, constructing a particle size identification model suitable for the suspension sample includes: Determining a training strategy for a recognition model according to the label value of the environment label, the training strategy including at least one of a training momentum, a regularization parameter, a batch size, a learning rate decay, and an initialization method, and determining a training parameter for the recognition model according to the network recognition parameter, the training parameter including at least one of a weight, a bias, and a learning rate; The recognition model that has completed the training strategy and training parameter configuration is determined as a particle size recognition model suitable for the suspension sample.

7. A device for identifying the particle size of nano metal powder, characterized in that: The device comprises: A sample acquisition unit, used to acquire a suspension sample of nano metal powder and identify a suspension medium of the suspension sample; A parameter acquisition unit is used to acquire network recognition parameters that match the suspension medium; when the network recognition parameters are generated, a standard suspension sample is constructed based on the suspension medium, and a plurality of different medium redundancies are added to the suspension medium to construct a plurality of different comparative suspension samples; wherein when the network recognition parameters trained based on the standard suspension sample are applied to the comparative suspension sample, the particle size prediction result obtained is different from the standard particle size prediction result; the medium redundancy represents the additional different types of suspension media added to the suspension medium; A label generating unit, used for determining the environmental characteristics of the suspension sample, analyzing the environmental characteristics, and generating an environmental label corresponding to the suspension sample; A model building unit, used to use the environmental label as an auxiliary parameter and the network identification parameter as a main parameter to build a particle size identification model suitable for the suspension sample; The particle size recognition unit is used to recognize the particle size of the nano metal powder in the suspension sample through the particle size recognition model.

8. The device according to claim 7, characterized in that The parameter acquisition unit further includes a comparison sample construction module, the comparison sample construction module is used to determine the standard medium coefficient of the suspension medium, and add medium redundancy on the basis of the standard medium coefficient to generate a plurality of different interference medium coefficients; The interfering suspension medium corresponding to each interfering medium coefficient is queried, and for any interfering suspension medium, an equal amount of nano-metal powder in the suspension sample is obtained, and a comparative suspension sample is constructed based on the equal amount of nano-metal powder and the interfering suspension medium.

9. The device according to claim 7 or 8, characterized in that The parameter acquisition unit is specifically used to construct a target sample set based on the standard suspension sample, and to construct a calibration sample set based on the comparative suspension sample, wherein the calibration sample set includes a plurality of comparative suspension samples with different medium redundancy; train a preset recognition model using the target sample set to obtain initial network parameters of the preset recognition model, wherein under the initial network parameters, the target sample set has a first particle size recognition result; process the comparative suspension samples in the calibration sample set using the preset recognition model with the initial network parameters to obtain a second particle size recognition result; cluster the second particle size recognition result, and determine whether the first particle size recognition result is located in the cluster cluster obtained by clustering, and if so, adjust the initial network parameters so that after processing the target sample set based on the adjusted network parameters, the first particle size recognition result is still obtained, and after clustering the third particle size recognition result obtained after processing the calibration sample set based on the adjusted network parameters, the first particle size recognition result is located outside the cluster cluster; determine the network parameters corresponding to the first particle size recognition result being located outside the cluster cluster as the network recognition parameters obtained by training based on the standard suspension sample.

10. The device according to claim 7, characterized in that The label generation unit is specifically used to collect temperature and humidity characteristics according to a preset sampling period, and arrange the collected temperature and humidity characteristics into a temperature sequence and a humidity sequence according to the collected timestamp; respectively extract the temperature characteristics of the temperature sequence and the humidity characteristics of the humidity sequence, and splice the temperature characteristics and the humidity characteristics into the environmental characteristics of the current suspension sample.

Images