Quality detection method and system for finished product of small molecule polysilicon pore-throat dredging agent

Abstract

The finished product quality detection method and system of the small molecule polysilicon pore throat dredging agent of the present application relate to the field of artificial intelligence technology. By setting SEM image acquisition parameters, the SEM image acquisition parameters include magnification and resolution, and SEM images of pores at different scales before and after dredging are acquired; multi-scale pore structure features before and after dredging are extracted from the SEM images; quantitative descriptors are calculated based on the SEM images, and the multi-scale pore structure features and the quantitative descriptors are subjected to feature splicing to obtain a comprehensive pore structure representation; taking the comprehensive pore structure representation as an input, based on the dredging effect prediction model, qualitative and quantitative indicators for the quality detection of the dredging agent are predicted, improving the accuracy of the finished product quality detection of the small molecule polysilicon pore throat dredging agent.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and particularly to a method and system for detecting the finished product quality of small molecule polysilicon pore throat dredging agents. Background Art

[0002] Polysilicon materials have a wide range of applications in many fields due to their unique properties, such as waterproof and anticorrosive coatings, biomedical materials, electronic packaging materials, etc. However, during the use of polysilicon materials, their pore throats may be blocked by impurities, affecting the performance of the materials. Using small molecule compounds as dredging agents can penetrate into the pores of polysilicon materials, react chemically with the blockages or dissolve them, thereby dredging the pore throats and restoring the performance of the materials. Compared with using macromolecule dredging agents, small molecule dredging agents have the advantages of small size and strong permeability.

[0003] Traditional methods for detecting the quality of dredging agents mainly rely on experimental tests and empirical judgments, lacking quantitative and intelligent means, resulting in insufficient detection accuracy. Summary of the Invention

[0004] This application aims to solve at least one of the technical problems in the related art to some extent. For this reason, an object of this application is to propose a method and system for detecting the finished product quality of small molecule polysilicon pore throat dredging agents, which improves the accuracy of detecting the finished product quality of small molecule polysilicon pore throat dredging agents.

[0005] One aspect of this application provides a method for detecting the finished product quality of small molecule polysilicon pore throat dredging agents, including:

[0006] Step S100: Set SEM image acquisition parameters, where the SEM image acquisition parameters include magnification and resolution, and acquire SEM images of pores at different scales before and after dredging;

[0007] Step S200: Extract multi-scale pore structure features before and after dredging according to the SEM images;

[0008] Step S300: Calculate quantitative descriptors according to the SEM images, splice the multi-scale pore structure features with the quantitative descriptors to obtain a comprehensive pore structure representation;

[0009] Step S400: Use the comprehensive pore structure representation as an input, and based on a dredging effect prediction model, predict the qualitative and quantitative indicators of dredging agent quality detection;

[0010] The specific method for setting the SEM image acquisition parameters is:

[0011] Step S110: Divide the pores of the sample before dredging and the sample after dredging into three scales according to the pore size and geometric shape, including micron-sized pores, mesopores, and macropores;

[0012] Step S120: Obtain the pore size range in the sample before dredging and the pore size range after dredging , where and are respectively the minimum and maximum pore sizes before dredging, and are respectively the minimum and maximum pore sizes after dredging;

[0013] Step S130: Set the value range of the magnification factor before dredging such that the pixel diameter of the minimum pore size before dredging in the SEM image is equal to the ratio of the minimum pore size before dredging to the actual size of each pixel , and the magnification factor before dredging is greater than or equal to the pixel diameter threshold of the minimum pore size before dredging in the SEM image ;

[0014] Step S140: Set the value range of the resolution before dredging , and the resolution before dredging is less than or equal to half of the minimum pore size before dredging;

[0015] Step S150: Set the value range of the magnification factor after dredging such that the pixel diameter of the minimum pore size after dredging in the SEM image is equal to the ratio of the minimum pore size after dredging to the actual size of each pixel , and the magnification factor after dredging is greater than or equal to the pixel diameter threshold of the minimum pore size after dredging in the SEM image ;

[0016] Step S160: Set the value range of the resolution after dredging , and the resolution after dredging is less than or equal to half of the minimum pore size after dredging;

[0017] Step S170: Preset the optimal pore pixel diameter , calculate the magnification factors for different pore sizes according to different pore sizes d and the optimal pore pixel diameter, and the magnification factors satisfy the corresponding value range of the magnification factor before dredging or the value range of the magnification factor after dredging ;

[0018] Step S180: Define the resolution according to different pore sizes , and the resolution satisfies the corresponding value range of the resolution before dredging or the value range of the resolution after dredging;

[0019] The specific method for extracting the multi-scale pore structure features before and after dredging from the SEM image is as follows:

[0020] Step S210: For micro-scale pores, extract the local texture features of the siloxane structure before and after dredging based on the Gabor filter , the parameters of the Gabor filter include the scale parameter , the orientation parameter and the frequency parameter ;

[0021] Step S220: For mesopores and macropores, extract their morphological features and region segmentation features by morphological operations and region segmentation algorithms;

[0022] Step S230: Fuse the local texture features of micro-scale pores with the morphological features and region segmentation features of mesopores and macropores by feature splicing to obtain multi-scale pore structure features ;

[0023] The setting method of the scale parameter is as follows: Obtain the resolution r and magnification m of the SEM image, as well as the pixel size range of the siloxane structure , then the corresponding physical size range is ; where , are the minimum and maximum values of the physical size range respectively;

[0024] Within the physical size range, select scale parameters by logarithmic equally spaced sampling, where is the th scale parameter, and the i-th scale parameter is ;

[0025] The specific method for extracting the local texture features of the siloxane structure is as follows:

[0026] For each set of parameters of the Gabor filter , calculate the convolution response of the Gabor filter on the SEM image to obtain a set of Gabor feature maps; Stitch the Gabor feature maps obtained from all the parameters of the Gabor filter into a long vector

[0027] , and the long vector is used as the local texture feature of the siloxane structure under micro-scale pores;

[0028] ​The specific method for calculating quantitative descriptors based on SEM images, stitching the multi-scale pore structure features with the quantitative descriptors, and obtaining the comprehensive pore structure representation is as follows:

[0029] Step S310: Calculate the quantitative descriptors D of the pores before and after dredging based on the SEM images. The quantitative descriptors include specific surface area , pore volume and pore size distribution . The BET method is used to calculate the specific surface area of the pores, the mercury intrusion method is used to measure the pore volume of the pores, and the BJH method is used to calculate the pore size distribution of the pores;

[0030] Step S320: Stitch the multi-scale pore structure features with the quantitative descriptors to obtain the comprehensive pore structure representation X;

[0031] The specific method for using the comprehensive pore structure representation as the input and predicting the qualitative and quantitative indicators of the quality inspection of the dredging agent based on the dredging effect prediction model is as follows:

[0032] Step S410: Obtain the quality inspection training samples of the finished dredging agent. The training samples include the comprehensive pore structure representations of the pores before and after dredging and the corresponding qualitative index labels and quantitative index labels;

[0033] Step S420: Design a multi-scale convolutional neural network. The multi-scale convolutional neural network contains n convolutional branches, and each convolutional branch is used to extract pore structure feature maps of different scales ;

[0034] Step S430: Design an attention mechanism module, introduce the attention mechanism, calculate the attention weights of the pore structure feature maps of each scale , obtain the weighted pore structure feature maps of each scale, and fuse the weighted pore structure feature maps to obtain the weighted fusion feature z;

[0035] Step S440: Design a pore structure evolution modeling module, use a long short-term memory network, take the weighted fusion feature at the current moment as the input, and output the hidden state at the current moment ;

[0036] Step S450: Construct a dredging effect prediction module, use the hidden state output by the pore structure evolution modeling module as the input through a fully connected layer, and use it to predict and output the qualitative and quantitative indicators of the quality inspection of the dredging agent;

[0037] Step S460: Define the cross-entropy loss function for the qualitative indicators and the mean square error loss function for the quantitative indicators , and weight and sum the cross-entropy loss function and the mean square error loss function to obtain the total loss function ;

[0038] Step S470: A dredging effect prediction model is composed of a multi-scale convolutional neural network, an attention mechanism module, a pore structure evolution modeling module, and a dredging effect prediction module. It is trained based on training samples with the minimization of the total loss function as the training objective. When the total loss function converges, the training ends. Based on the trained dredging effect prediction model, the qualitative and quantitative indicators of the quality inspection of the new finished dredging agent are predicted.

[0039] One aspect of the present application provides a quality inspection system for the finished product of a small molecule polysilicon pore throat dredging agent, including:

[0040] An SEM image acquisition module for setting SEM image acquisition parameters, where the SEM image acquisition parameters include magnification and resolution, and acquiring SEM images of pores at different scales before and after dredging;

[0041] A multi-scale feature extraction module for extracting multi-scale pore structure features before and after dredging according to the SEM images;

[0042] A comprehensive feature representation module for calculating quantitative descriptors according to the SEM images, and performing feature splicing on the multi-scale pore structure features and the quantitative descriptors to obtain a comprehensive pore structure representation;

[0043] A quality inspection output module for using the comprehensive pore structure representation as an input, and predicting the qualitative and quantitative indicators of the quality inspection of the dredging agent based on the dredging effect prediction model.

[0044] The quality inspection method and system for the finished product of the small molecule polysilicon pore throat dredging agent proposed in the present application have the following advantages compared with the prior art:

[0045] By introducing multi-scale pore structure features and a deep learning model, the present application extracts multi-scale features of micron-scale pores, mesopores, and macropores before and after dredging, comprehensively depicts the changes in pore structures at different scales, provides richer and more comprehensive information for comprehensively evaluating the dredging effect of the dredging agent, and improves the detection accuracy.

[0046] For micron-scale pores, the present application uses a Gabor filter to extract local texture features, which can effectively capture the changes in the siloxane structure before and after dredging, reflect the influence of the dredging agent on microscopic pores. By adjusting the scale, direction, and frequency parameters, the Gabor filter can adaptively extract texture features of different sizes and directions, improving the flexibility and adaptability of feature extraction.

[0047] By performing feature fusion on multi-scale features, the present application comprehensively depicts the changes in pore structures before and after dredging, improving the expression ability and generalization performance of the prediction model.

[0048] This application automatically learns key features at different scales through a multi-scale convolution structure, without the need for manual feature design, improving the adaptability and intelligence of feature extraction. By introducing an attention mechanism, it dynamically focuses on key regions and features at different scales according to the differences in dredging effects, enhancing the adaptability and robustness of the prediction model. The long short-term memory network is used to capture the temporal dependence relationship of pore morphology and quantitative indicators before and after dredging, making full use of the dynamic information during the dredging process.

[0049] Through multi-scale pore structure feature extraction and deep learning model construction, this application effectively solves the problems in traditional quality detection methods of dredging agents, such as the lack of quantitative and intelligent means and insufficient detection accuracy. It comprehensively depicts the changes in pore structures at different scales, provides rich and comprehensive information, greatly improves the accuracy, efficiency, and intelligence level of dredging agent quality detection, and provides reliable technical support for the research and development and application of dredging agents. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 It is a method flow chart of the finished product quality detection method for the small molecule polysilicon pore throat dredging agent provided by this application;

[0051] Figure 2 It is a method flow chart of the acquisition method for multi-scale pore structure features provided by this application;

[0052] Figure 3 It is a method flow chart of the acquisition method for the comprehensive pore structure representation provided by this application;

[0053] Figure 4 It is a functional module diagram of the finished product quality detection system for the small molecule polysilicon pore throat dredging agent provided by this application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0054] To better understand this application, more detailed descriptions of various aspects of this application will be made with reference to the accompanying drawings. It should be understood that these detailed descriptions are only descriptions of the exemplary embodiments of this application and do not limit the scope of this application in any way. Throughout the specification, the same reference numerals refer to the same elements. The expression "and / or" includes any and all combinations of one or more of the associated listed items.

[0055] In the accompanying drawings, for the sake of clarity, the sizes, dimensions, and shapes of the elements have been slightly adjusted. The accompanying drawings are for illustrative purposes only and are not drawn to an exact scale. As used herein, terms such as "substantially", "about", and similar terms are used as terms of approximation and not as terms of degree, and are intended to account for the inherent deviations in measured or calculated values that would be recognized by a person of ordinary skill in the art. Additionally, in this application, the order in which the steps of the processes are described does not necessarily represent the order in which these processes occur in actual operation, unless otherwise explicitly specified or derivable from the context.

[0056] It should also be understood that expressions such as "comprising", "including", "having", "containing", and / or "including having" are open-ended rather than closed-ended expressions in this specification, which means that the stated features, elements, and / or components exist, but do not exclude the existence of one or more other features, elements, components, and / or combinations thereof. In addition, when an expression such as "at least one of..." appears after a list of listed features, it modifies the entire list of features rather than just individual elements in the list. Furthermore, when describing the embodiments of this application, the use of "may" indicates "one or more embodiments of this application". And the term "exemplary" is intended to refer to an example or illustration.

[0057] Unless otherwise defined, all terms used herein (including engineering and scientific terms) have the same meaning as commonly understood by a person of ordinary skill in the art to which this application belongs. It should also be understood that, unless explicitly stated otherwise in this application, words defined in commonly used dictionaries should be interpreted as having a meaning consistent with their meaning in the context of the relevant art and should not be interpreted in an idealized or overly formal sense.

[0058] It should be noted that, without conflict, the embodiments in this application and the features in the embodiments can be combined with each other. The following will describe this application in detail with reference to the accompanying drawings and in combination with the embodiments.

[0059] Example 1

[0060] As Figure 1 shown, the method for detecting the finished product quality of the small molecule polysilicon pore throat dredging agent provided by this application includes:

[0061] Step S100: Set the SEM image acquisition parameters, where the SEM image acquisition parameters include the magnification and resolution, and acquire the SEM images of pores at different scales before and after dredging;

[0062] The specific method for setting the SEM image acquisition parameters is:

[0063] Step S110: Divide the pores of the sample before dredging and the sample after dredging into three scales according to the pore size and geometric shape, including micro-scale pores, mesopores, and macropores;

[0064] The division rules for the micro-scale pores, mesopores, and macropores are as follows:

[0065] Pores with a pore size greater than or equal to 1 μm and an irregular geometric shape are labeled as micro-scale pores; the micro-scale pores can be directly observed in the SEM image, with obvious morphological characteristics and irregular geometric shapes, such as slit-shaped, fissure-shaped, and dendritic;

[0066] Pores with a pore size greater than or equal to 2 nm and less than or equal to 50 nm are labeled as mesopores; the mesopores usually require a higher magnification to be resolved in the SEM image, and the morphological characteristics are more refined than those of the micro-scale pores;

[0067] Pores with a pore size greater than 50 nm and a regular geometric shape are labeled as macropores; the macropores have regular geometric shapes, such as spherical, cylindrical, honeycomb-shaped, etc., the orientation and arrangement of the pores are relatively orderly, and the pore walls are relatively smooth;

[0068] Step S120: Obtain the pore size range of the pores in the sample before dredging and the pore size range after dredging , where and are the minimum and maximum pore sizes before dredging, respectively, and are the minimum and maximum pore sizes after dredging, respectively;

[0069] Step S130: Set the value range of the magnification before dredging, such that the pixel diameter of the minimum pore size before dredging in the SEM image is equal to the ratio of the minimum pore size before dredging to the actual size of each pixel, and the magnification before dredging is greater than or equal to the pixel diameter threshold of the minimum pore size before dredging in the SEM image;

[0070] The pixel diameter threshold of the minimum pore size before dredging in the SEM image is preset by those skilled in the art according to experience; preferably, the value range of the pixel diameter threshold is 5-10;

[0071] The pixel diameter of the minimum pore size before dredging satisfies: ;

[0072] From the above formula, it can be obtained that the value range of the magnification before dredging satisfies: ;

[0073] Step S140: Set the resolution before dredging within a certain range, where the resolution before dredging is less than or equal to half of the minimum pore diameter before dredging;

[0074] The resolution before dredging is: ;

[0075] Step S150: Set the magnification after dredging within a certain range such that the pixel diameter of the minimum pore diameter after dredging in the SEM image is equal to the ratio of the minimum pore diameter after dredging to the actual size of each pixel , and the magnification after dredging is greater than or equal to the pixel diameter threshold of the minimum pore diameter after dredging in the SEM image ;

[0076] The pixel diameter threshold of the minimum pore diameter after dredging in the SEM image is preset by those skilled in the art according to experience; preferably, the value range of the pixel diameter threshold is 10 - 20;

[0077] The pixel diameter of the minimum pore diameter after dredging satisfies: ;

[0078] From the above formula, the value range of the magnification after dredging satisfies: ;

[0079] Step S160: Set the resolution after dredging within a certain range, where the resolution after dredging is less than or equal to half of the minimum pore diameter after dredging;

[0080] The resolution after dredging is: ;

[0081] Step S170: Preset the optimal pore pixel diameter , calculate the magnification of different pore diameters according to different pore diameters d and the optimal pore pixel diameter, and the magnification satisfies the corresponding value range of the magnification before dredging or the value range of the magnification after dredging ;

[0082] The calculation formula for the magnification of different pore diameters is: ;

[0083] The value of the optimal pore pixel diameter is set by those skilled in the art according to the pore diameter and imaging requirements. Preferably, the optimal pore pixel diameter The value range of [] is 10 to 50;

[0084] The pore diameter mentioned above is smaller, the magnification is larger, so as to obtain sufficient pore detail information;

[0085] Step S180: Define the resolution according to different pore diameters and the resolution satisfies the value range of the pre-unclogging resolution or the post-unclogging resolution corresponding thereto;

[0086] The resolution mentioned above has the following calculation formula: where q is the resolution coefficient;

[0087] Preferably, the value of the resolution coefficient is 2 to 5, and the specific value is selected according to the imaging requirements;

[0088] The smaller the pore diameter, the lower the resolution, so as to distinguish the fine structure of the pores.

[0089] The resolution refers to the minimum resolvable distance of the instrument. When the resolution coefficient is set to 2, for a pore of one micron, the required resolution is 0.5, indicating that the smaller the pore, in order to ensure the same resolution coefficient, the instrument needs to have a higher resolution ability, that is, the R value is smaller.

[0090] The specific method for collecting SEM images of pores of different scales before and after unclogging is as follows:

[0091] Collect SEM images of pores of different scales according to the corresponding magnification and resolution. Perform wavelet transform denoising on the SEM images of micron-scale pores and adjust the contrast;

[0092] For the SEM images of mesopores and macropores, smooth the pore regions to enhance the contrast between the pore regions and the matrix;

[0093] The above steps determine the magnification and resolution adaptively by setting the SEM image acquisition parameters according to the pore diameter size ranges of pores of different scales before and after unclogging, ensuring the acquisition of high-quality pore structure images, laying a foundation for subsequent feature extraction and analysis.

[0094] Step S200: Extract multi-scale pore structure features before and after unclogging according to the SEM images;

[0095] As Figure 2 shown, the specific method for extracting multi-scale pore structure features before and after unclogging according to the SEM images is as follows:

[0096] Step S210: For micro-scale pores, extract the local texture features of the siloxane structure before and after dredging based on the Gabor filter. , the parameters of the Gabor filter include the scale parameter , the orientation parameter and the frequency parameter ;

[0097] The setting method of the scale parameter is as follows: Obtain the resolution r and magnification m of the SEM image, as well as the pixel size range of the siloxane structure , then the corresponding physical size range is ; where , are the minimum and maximum values of the physical size range respectively;

[0098] The pixel size range of the siloxane structure is obtained through experimental measurement;

[0099] Within the physical size range, select scale parameters using a logarithmic equally spaced sampling method, where is the th scale parameter, and the i-th scale parameter is ;

[0100] The calculation formula for the scale parameter is: , where ;

[0101] The scale parameter is used to control the size of the filter and is adapted to the size of the siloxane structure.

[0102] The setting method of the orientation parameter is as follows: Select orientation parameter values using an equally spaced sampling method, where represents the th orientation parameter;

[0103] The calculation formula for the orientation parameter is: , where ;

[0104] The orientation parameter is used to control the direction of the filter, and common orientation parameters include 4, 6, 8, etc., to capture texture features in different directions;

[0105] The setting method of the frequency parameter is as follows: Obtain the frequency range of the siloxane structure texture;

[0106] The method for obtaining the frequency range of the siloxane structural texture is as follows: perform a Fourier transform on the SEM image, analyze the spectral characteristics of the siloxane structural texture, and determine the frequency range. ;

[0107] Within the said frequency range, select frequency parameters by using a logarithmic equally spaced sampling method ; where is the th frequency parameter;

[0108] The calculation formula of the said frequency parameter is: , where .

[0109] In summary, the parameter setting of the Gabor filter can be represented as a triple .

[0110] The specific method for extracting the local texture features of the siloxane structure is as follows:

[0111] For each set of parameters of the Gabor filter , calculate the convolution response of the Gabor filter on the SEM image to obtain a set of Gabor feature maps;

[0112] Stitch the Gabor feature maps obtained from all the parameters of the Gabor filter into a long vector , and the said long vector serves as the local texture features of the siloxane structure under the micron-sized pores;

[0113] The calculation formula for the convolution response of the Gabor filter on the SEM image is: , where (x, y) represents the pixel coordinates in the Gabor feature map, represents the pixel coordinates on the SEM image, is the mother function of the Gabor filter;

[0114] The calculation formula for the mother function of the Gabor filter is: , where j* is the imaginary unit, and s, , f are the scale, orientation, and frequency respectively;

[0115] Step S220: For mesopores and macropores, extract their morphological features and region segmentation features ;

[0116] The said morphological features include but are not limited to pore size, geometric shape, and pore orientation;

[0117] The regional segmentation features include, but are not limited to, the number of pores, pore connectivity, pore distribution, and pore area;

[0118] The morphological operations include erosion operation, dilation operation, opening operation, and closing operation; the erosion operation is used to remove the noise and small protrusions at the pore edges of mesopores and macropores, the dilation operation is used to fill the small voids inside the pores to make the pore shape more complete, and the opening operation and closing operation are used to smooth the pore boundaries and eliminate small connected regions;

[0119] The regional segmentation algorithm uses a threshold-based segmentation algorithm. According to the gray-scale difference between the pores of mesopores and macropores and the matrix of the sample material to be dredged, a preset threshold is used to segment the SEM image into two parts: pores and matrix;

[0120] The above steps extract multi-scale pore structure features, adopt different feature extraction methods for pores of different scales. For example, for micron-scale pores, Gabor filtering is used to extract texture features, and for mesopores and macropores, morphological operations and regional segmentation are used to extract morphological and regional features, and feature fusion is performed to comprehensively characterize the changes in the pore structure before and after dredging.

[0121] Step S230: Fuse the local texture features of micron-scale pores with the morphological features and regional segmentation features of mesopores and macropores by feature stitching to obtain multi-scale pore structure features .

[0122] Step S300: Calculate quantitative descriptors based on the SEM image, and perform feature stitching on the multi-scale pore structure features and the quantitative descriptors to obtain a comprehensive pore structure representation;

[0123] Such as Figure 3 As shown, the specific method for calculating quantitative descriptors based on the SEM image, performing feature stitching on the multi-scale pore structure features and the quantitative descriptors to obtain a comprehensive pore structure representation is:

[0124] Step S310: Calculate the quantitative descriptors D of the pores before and after dredging based on the SEM image. The quantitative descriptors include specific surface area , pore volume and pore size distribution . The BET method is used to calculate the specific surface area of the pores, the mercury intrusion method is used to measure the pore volume of the pores, and the BJH method is used to calculate the pore size distribution of the pores;

[0125] The quantitative descriptors can be expressed as: ;

[0126] The calculation method of the specific surface area is: using the BET equation, and the BET equation is , by linear fitting for the relationship, the monolayer adsorption capacity and the BET constant value are obtained, and then the specific surface area is calculated; where, v is the volume of the adsorbed gas, and are the equilibrium pressure and the saturated vapor pressure respectively, is the monolayer adsorption capacity, and c is the BET constant;

[0127] The calculation formula for the specific surface area is: ; where, is the Avogadro constant, S is the cross-sectional area of the adsorbate molecule, and V is the volume of the sample to be dredged;

[0128] The saturated vapor pressure is obtained by looking up the table according to the adsorbate and temperature;

[0129] The volume of the adsorbed gas and the equilibrium pressure are obtained through experimental measurement;

[0130] The cross-sectional area of the adsorbate molecule is obtained by looking up the table according to the type of adsorbate. The cross-sectional area of the commonly used adsorbate nitrogen is 0.162 nm 2 ;

[0131] The volume of the sample to be dredged is obtained by the fluid displacement method.

[0132] The calculation method for the pore volume is: using the mercury intrusion method, and the mercury intrusion method is based on the Washburn equation. By measuring the mercury column volume at different pressures, the pore volume is obtained;

[0133] The equation of the Washburn equation is: ; where, P is the mercury intrusion pressure, is the surface tension of mercury, is the contact angle between mercury and the pore wall, is the pore size;

[0134] The calculation formula for the pore volume is: ; where, is the volume distribution function of pores with a pore size of ;

[0135] The mercury intrusion pressure and the volume distribution function of pores are obtained through mercury intrusion experiments;

[0136] The contact angle between mercury and the pore wall is related to the pore wall material. Preferably, the value of the contact angle is generally 130° - 140°;

[0137] The calculation method of the pore size distribution is as follows: The BJH method is adopted, and the BJH method is based on the Kelvin equation to calculate the pore size distribution of pores;

[0138] The equation of the Kelvin equation is: ; where is the molar volume of the adsorbate, is the ideal gas constant, is the absolute temperature, is the surface tension of the adsorbate, is the contact angle between the adsorbate and the pore wall;

[0139] The calculation formula of the pore size distribution of the pores is: ;

[0140] The surface tension and molar volume of the adsorbate are obtained by looking up the table according to the type of the adsorbate;

[0141] The contact angle between the adsorbate and the pore wall is set to 0;

[0142] Step S320: Perform feature splicing on the multi-scale pore structure characteristics and quantitative descriptors to obtain the comprehensive pore structure representation X;

[0143] The comprehensive pore structure representation is: ;

[0144] By combining the quantitative descriptors reflecting the physical properties of the pores with the multi-scale features reflecting the pore morphology and distribution in the above steps, a comprehensive pore structure representation is formed, enhancing the expression ability and complementarity of the features.

[0145] Step S400: Using the comprehensive pore structure representation as the input, based on the dredging effect prediction model, predict the qualitative and quantitative indicators for the quality inspection of the dredging agent;

[0146] The specific method of using the comprehensive pore structure representation as the input and predicting the qualitative and quantitative indicators for the quality inspection of the dredging agent based on the dredging effect prediction model is:

[0147] Step S410: Obtain the quality inspection training samples of the finished dredging agent, and the training samples include the comprehensive pore structure representations of the pores before and after dredging and the corresponding qualitative index labels and quantitative index labels;

[0148] The qualitative index refers to the quality grade index qualitatively evaluated by those skilled in the art;

[0149] The quantitative index refers to the flow guiding ability of the small molecule polysilicon pore throat after dredging obtained through experimental measurement, which is used to evaluate the improvement degree of the dredging effect;

[0150] Step S420: Design a multi-scale convolutional neural network, which includes n convolutional branches, and each convolutional branch is used to extract pore structure feature maps of different scales. ;

[0151] The multi-scale convolutional neural network consists of multiple parallel convolutional branches, and each convolutional branch uses a different convolutional kernel size to capture features of different scales.

[0152] The expression of the pore structure feature map is: , where is the pore structure feature map of the a-th scale, is the mapping function of the convolutional branch of the a-th scale, X represents the input comprehensive pore structure representation, represents the convolutional kernel weight of the a-th scale, represents the convolutional kernel bias of the a-th scale, is the activation function;

[0153] The output of the multi-scale convolutional neural network is denoted as , where is the pore structure feature map of the a-th scale;

[0154] Step S430: Design an attention mechanism module, introduce the attention mechanism, calculate the attention weights of the pore structure feature maps of each scale , obtain the weighted pore structure feature maps of each scale, and fuse the weighted pore structure feature maps to obtain the weighted fusion feature z;

[0155] The calculation formula of the attention weight is: , where is the transpose of the attention weight vector of the a-th scale , is the transpose of the attention weight vector of the g-th scale , is the pore structure feature map of the g-th scale;

[0156] The calculation formula of the weighted fusion feature is: ;

[0157] Step S440: Design a pore structure evolution modeling module, use a long short-term memory network, take the weighted fusion feature at the current moment as the input, and output the hidden state at the current moment;

[0158] The state transition formula of the long short-term memory network is:

[0159] ;

[0160] ;

[0161] ;

[0162] ;

[0163] ;

[0164] ;

[0165] in, 、 、 、 They are input gate, forget gate, output gate, and candidate memory unit state respectively. 、 、 、 are the weight matrices of the input gate, forget gate, output gate, and candidate memory unit states, respectively. 、 、 、 are the bias vectors of the input gate, forget gate, output gate, and candidate memory unit states, respectively. is the memory unit state, 、 are the hidden states of the current moment and the previous moment respectively, is the weighted fusion feature at the current moment, is the sigmoid activation function, is element-wise multiplication, is the hyperbolic tangent activation function;

[0166] Step S450: Constructing a dredging effect prediction module, using the hidden state output by the pore structure evolution modeling module as input through a fully connected layer, to predict and output qualitative and quantitative indicators of the dredging agent quality test;

[0167] The prediction formula of the qualitative indicator is: ,in, and is the weight matrix and bias vector of the fully connected layer used to predict qualitative indicators;

[0168] The prediction formula of the quantitative indicator is: ,in, and is the weight matrix and bias vector of the fully connected layer used to predict quantitative indicators;

[0169] Step S460: Define the cross entropy loss function of qualitative indicators And the mean square error loss function of quantitative indicators The cross-entropy loss function and the mean squared error loss function are weighted and summed to obtain the total loss function;

[0170] The calculation formula of the total loss function is: where and are the weight coefficients of the cross-entropy loss function and the mean squared error loss function respectively;

[0171] The weight coefficients of the cross-entropy loss function and the mean squared error loss function are set by those skilled in the art according to experience, and The sum of them is 1;

[0172] The calculation formula of the cross-entropy loss function is: where and are the true qualitative index label and the predicted qualitative index of the ni-th training sample respectively, and N represents the number of training samples;

[0173] The calculation formula of the mean squared error loss function is: where and are the true quantitative index label and the predicted quantitative index of the ni-th training sample respectively;

[0174] Step S470: A dredging effect prediction model is composed of a multi-scale convolutional neural network, an attention mechanism module, a pore structure evolution modeling module, and a dredging effect prediction module. It is trained based on training samples, with minimizing the total loss function as the training objective. When the total loss function converges, the training ends. Based on the trained dredging effect prediction model, the qualitative index and the quantitative index of the quality inspection of the new dredging agent finished product are predicted.

[0175] The above steps extract feature maps of different scales by using a multi-scale convolutional neural network, introduce an attention mechanism to adaptively adjust the importance of each scale feature, use a long short-term memory network to model the dynamic evolution process of the pore structure, and predict the qualitative index and the quantitative index simultaneously through multi-task learning, improving the prediction performance and generalization ability of the model.

[0176] Embodiment 2

[0177] As Figure 4 shown, the quality inspection system for the finished product of the small molecule polysilicon pore throat dredging agent provided by this application includes:

[0178] An SEM image acquisition module, which is used to set SEM image acquisition parameters. The SEM image acquisition parameters include magnification and resolution, and collect SEM images of pores at different scales before and after dredging;

[0179] A multi-scale feature extraction module for extracting multi-scale pore structure features before and after dredging based on SEM images;

[0180] A comprehensive feature representation module for calculating quantitative descriptors based on SEM images, and performing feature splicing on the multi-scale pore structure features and the quantitative descriptors to obtain a comprehensive pore structure representation;

[0181] A quality detection output module for taking the comprehensive pore structure representation as input and predicting the qualitative and quantitative indicators of the quality detection of the dredging agent based on the dredging effect prediction model.

[0182] In the above technical solutions provided in the embodiments of the present application, the parts that are consistent with the corresponding technical solutions in the prior art in terms of implementation principles are not described in detail to avoid excessive elaboration.

[0183] As described above in the specific embodiments, the purpose, technical solutions and beneficial effects of the present invention are further described in detail. It should be understood that the above are only specific embodiments of the present invention and are not used to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included within the protection scope of the present invention.

Claims

1. A method for detecting the finished product quality of a small molecule polysilicon pore throat dredging agent, characterized in that, Including: Set SEM image acquisition parameters, where the SEM image acquisition parameters include magnification and resolution, and acquire SEM images of pores at different scales before and after dredging; Extract multi-scale pore structure features before and after dredging from the SEM images; Calculate quantitative descriptors, splice the multi-scale pore structure features with the quantitative descriptors to obtain a comprehensive pore structure representation; Using the comprehensive pore structure representation as input, based on the dredging effect prediction model, predict the qualitative and quantitative indicators for the quality inspection of the dredging agent; The specific method for extracting multi-scale pore structure features before and after dredging from the SEM images is: For micron-sized pores, local texture features of the siloxane structure before and after dredging are extracted based on the Gabor filter , the parameters of the Gabor filter include the scale parameter , the orientation parameter and the frequency parameter ; For mesopores and macropores, their morphological features and regional segmentation features are extracted through morphological operations and regional segmentation algorithms and regional segmentation features ; Fuse the local texture features of microscale pores with the morphological features and regional segmentation features of mesopores and macropores by feature splicing to obtain multi-scale pore structure features ; The quantitative descriptors include specific surface area , pore volume and pore size distribution ; The dredging effect prediction model consists of a multi-scale convolutional neural network, an attention mechanism module, a pore structure evolution modeling module, and a dredging effect prediction module.

2. The finished product quality inspection method of the small molecule polysilicon pore throat dredging agent according to claim 1, characterized in that, The specific method for setting the SEM image acquisition parameters includes: Divide the pores of the sample before dredging and the sample after dredging into three scales according to pore size and geometric shape, including micro-scale pores, mesopores, and macropores; Obtain the pore size range of pores in the sample before dredging and the pore size range after dredging , where and are the minimum and maximum pore sizes before dredging respectively, and are the minimum and maximum pore sizes after dredging respectively; Set the magnification before dredging within a certain range so that the pixel diameter of the smallest pore diameter before dredging in the SEM image is equal to the ratio of the smallest pore diameter before dredging to the actual size of each pixel , and the magnification before dredging is greater than or equal to the pixel diameter threshold of the smallest pore diameter before dredging in the SEM image ; Set the resolution before dredging whose value range is such that the resolution before dredging is less than or equal to half of the minimum aperture before dredging; Set the magnification factor after dredging within a certain range so that the pixel diameter of the smallest pore diameter after dredging in the SEM image is equal to the ratio of the smallest pore diameter after dredging to the actual size of each pixel , and the magnification factor after dredging is greater than or equal to the pixel diameter threshold of the smallest pore diameter after dredging in the SEM image ; Set the resolution after dredging The value range is such that the resolution after dredging is less than or equal to half of the minimum aperture after dredging.

3. The finished product quality inspection method of the small molecule polysilicon pore throat dredging agent according to claim 2, characterized in that, The specific method for setting the SEM image acquisition parameters also includes: Preset optimal pore pixel diameter , calculate the magnification of different pore sizes according to different pore diameters d and the optimal pore pixel diameter, and the magnification satisfies the corresponding magnification before dredging in the value range or the magnification after dredging in the value range; Define the resolution according to different pore sizes , and the resolution satisfies the value range of the resolution before dredging or the value range of the resolution after dredging corresponding thereto.

4. The finished product quality inspection method of the small molecule polysilicon pore throat dredging agent according to claim 3, characterized in that The setting method of the scale parameter is as follows: Obtain the resolution r and magnification m of the SEM image, as well as the pixel size range of the siloxane structure , then the corresponding physical size range is ; where , are the minimum and maximum values of the physical size range respectively; Within the said physical size range, logarithmically equally-spaced sampling is adopted to select scale parameters , where is the -th scale parameter, and the i-th scale parameter is .

5. The finished product quality inspection method of the small molecule polysilicon pore throat dredging agent according to claim 4, characterized in that, The specific method for calculating quantitative descriptors, splicing the multi-scale pore structure features with the quantitative descriptors to obtain a comprehensive pore structure representation is: Calculate the quantitative descriptor D of the pores before and after dredging, calculate the specific surface area of the pores using the BET method, measure the pore volume of the pores using the mercury intrusion method, and calculate the pore size distribution of the pores using the BJH method; Splice the multi-scale pore structure features with the quantitative descriptors to obtain a comprehensive pore structure representation X.

6. The finished product quality inspection method of the small molecule polysilicon pore throat dredging agent according to claim 5, characterized in that The specific method for using the comprehensive pore structure representation as input, based on the dredging effect prediction model, to predict the qualitative and quantitative indicators for the quality inspection of the dredging agent includes: Obtain the quality inspection training samples of the finished dredging agent, where the training samples include the comprehensive pore structure representations of the pores before and after dredging and the corresponding qualitative indicator labels and quantitative indicator labels; Design a multi-scale convolutional neural network, where the multi-scale convolutional neural network includes n convolutional branches, and each convolutional branch is used to extract pore structure feature maps of different scales ; Design an attention mechanism module, introduce the attention mechanism, and calculate the attention weights of the pore structure feature maps at each scale , obtain the weighted pore structure feature maps at each scale, fuse the weighted pore structure feature maps, and obtain the weighted fusion feature z; Design a pore structure evolution modeling module, which uses a long short-term memory network and takes the weighted fusion feature at the current moment as the input to output the hidden state at the current moment ; Construct a dredging effect prediction module, using the hidden state output by the pore structure evolution modeling module as input through a fully connected layer, for predicting the qualitative and quantitative indicators for the quality inspection of the dredging agent.

7. The finished product quality inspection method of the small molecule polysilicon pore throat dredging agent according to claim 6, characterized in that, The specific method for using the comprehensive pore structure representation as input, based on the dredging effect prediction model, to predict the qualitative and quantitative indicators for the quality inspection of the dredging agent also includes: Define the cross-entropy loss function for qualitative indicators and the mean squared error loss function for quantitative indicators , and obtain the total loss function by weighted summation of the cross-entropy loss function and the mean squared error loss function ; Train based on the training samples, with minimizing the total loss function as the training objective, and end the training when the total loss function converges. Predict the qualitative and quantitative indicators for the quality inspection of the new finished dredging agent based on the trained dredging effect prediction model.

8. A finished product quality inspection system for small molecule polysilicon pore throat dredging agent, which is used to implement the finished product quality inspection method for small molecule polysilicon pore throat dredging agent described in any one of claims 1-7, characterized in that, Including: An SEM image acquisition module, used to set SEM image acquisition parameters, where the SEM image acquisition parameters include magnification and resolution, and acquire SEM images of pores at different scales before and after dredging; A multi-scale feature extraction module, used to extract multi-scale pore structure features before and after dredging from the SEM images; A comprehensive feature representation module, used to calculate quantitative descriptors, splice the multi-scale pore structure features with the quantitative descriptors to obtain a comprehensive pore structure representation; A quality inspection output module, which uses the comprehensive pore structure representation as input and predicts the qualitative and quantitative indicators of the dredging agent quality inspection based on the dredging effect prediction model.

Images