High-purity quartz sand quality on-line detection method, device, equipment and medium

Abstract

The invention relates to a high-purity quartz sand quality on-line detection method, device, equipment and medium, and belongs to the technical field of image processing.The high-purity quartz sand quality on-line detection method comprises the steps that image data of quartz sand in each purification production link is obtained; performing segmentation standardization processing on the image data based on the trained image data segmentation standardization model to obtain quartz sand preliminary parameter data of each purification production link; matching the quartz sand preliminary parameter data of each purification production link with the standard sample data set based on a trained matching model, and determining standard sample data corresponding to the quartz sand preliminary parameter data; and inputting the quartz sand distribution characteristic data and the standard sample data corresponding to the quartz sand preliminary parameter data into a trained quartz sand quality detection and evaluation model to obtain a high-purity quartz sand quality evaluation result of each purification production link, so that the quartz sand quality detection accuracy is improved.

Description

Technical Field

[0001] This invention relates to the field of image processing technology, and in particular to a method, apparatus, equipment, and medium for online detection of the quality of high-purity quartz sand. Background Technology

[0002] High-purity quartz sand is a high-end mineral raw material obtained from natural quartz ore through a series of sorting and impurity removal processes. Its purity is usually above 99.95% or higher, and it is widely used in high-end optical devices, laser devices, optical communication, semiconductors, photovoltaics, microelectronics and other fields.

[0003] The production process of high-purity quartz sand has a significant impact on its quality stability. Through a series of physical and chemical purification technologies, impurities can be effectively removed and purity improved.

[0004] In the sand making process, fluctuations in the quality of high-purity quartz sand directly affect the subsequent product processes and product qualification rates, thus impacting sand sales and causing significant production losses. In the high-purity quartz sand production process, quartz particle size distribution is usually determined by manual sieving; the detection and identification of quartz particle shape, internal inclusion distribution, and mixed contaminants are usually obtained through manual sampling and microscopic photography to obtain qualitative evaluation results. Given that current high-purity quartz sand production lines mostly adopt open, intermittent, planar, and modular production methods, problems such as high labor costs, high labor intensity, low monitoring efficiency, and large detection errors occur in the quality monitoring process, leading to unstable quality of high-purity quartz sand. Summary of the Invention

[0005] In view of this, it is necessary to provide a method, device, equipment and storage medium for online detection of high-purity quartz sand quality, in order to solve the technical problem of low accuracy of quartz sand quality detection during the purification and production of high-purity quartz sand.

[0006] To address the aforementioned problems, in a first aspect, the present invention provides an online method for detecting the quality of high-purity quartz sand, comprising: Acquire image data of quartz sand in each purification and production stage; The image data is segmented and standardized based on the trained image data segmentation and standardization model to obtain the quartz sand distribution feature data of each purification and production stage. The quartz sand distribution feature data is then combined with the preset production conditions to obtain the preliminary parameter data of quartz sand for each purification and production stage. Based on the trained matching model, the preliminary parameter data of quartz sand in each purification and production stage is matched with the standard sample dataset to determine the standard sample data corresponding to the preliminary parameter data of quartz sand. The standard sample data corresponding to the quartz sand distribution characteristics data and the preliminary parameter data of the quartz sand are input into the trained quartz sand quality detection and evaluation model to obtain the high-purity quartz sand quality evaluation results for each purification and production stage.

[0007] In one possible implementation, the purification production process includes physical beneficiation and chemical beneficiation. The physical beneficiation includes grinding, magnetic separation, and flotation. The chemical beneficiation includes hot-pressing acid leaching and chlorination roasting. The image data of the quartz sand includes image data of quartz sand particle size, particle shape, inclusions, and contaminants.

[0008] In one possible implementation, the image data segmentation and standardization model includes a depthwise separable convolutional module, an inverse residual module, a linear bottleneck module, a CA attention module, and an activation function; the preset production conditions include physical beneficiation and chemical beneficiation conditions; the image data is segmented and standardized based on the trained image data segmentation and standardization model to obtain quartz sand distribution feature data corresponding to each purification production stage, and the quartz sand distribution feature data is merged with the preset production conditions to obtain preliminary parameter data of quartz sand for each purification production stage, including: The image data is preprocessed; The preprocessed image data is input into a trained image data segmentation and standardization model, and the preprocessed image data is extracted using a depthwise separable convolution module to obtain a feature map. The feature map is linearly reduced in dimension by using the inverse residual module and the linear bottleneck module to obtain a low-dimensional feature map. After performing global average pooling on the low-dimensional feature map, the pooled feature map is decomposed through the CA attention module to obtain the quartz sand distribution feature data corresponding to each purification and production stage. The quartz sand distribution characteristic data is combined with preset production conditions by using an activation function to obtain preliminary parameter data of quartz sand for each purification production stage.

[0009] In one possible implementation, the quartz sand distribution characteristic data includes quartz sand particle size distribution characteristics, quartz sand particle shape distribution characteristics, quartz sand inclusion content characteristics, and quartz sand contaminant content characteristics.

[0010] In one possible implementation, the matching model includes a feature fusion module; the matching of preliminary parameter data of quartz sand from each purification production stage and a standard sample dataset based on the trained matching model to determine the standard sample data corresponding to the preliminary parameter data of the quartz sand includes: The preliminary parameter data of the quartz sand and the standard sample dataset were normalized. The cosine similarity algorithm of the feature fusion module is used to calculate the similarity between the normalized preliminary parameter data of quartz sand and the standard sample dataset, and generate a similarity score matrix. The similarity score matrix is ​​evaluated based on a preset threshold to determine the matching result. Based on the matching results, standard sample data corresponding to the preliminary parameter data of the quartz sand are determined, wherein the matching results are evaluated by accuracy, precision, and recall.

[0011] In one possible implementation, the accuracy rate is: , in, For accuracy, The actual value is positive, and the predicted value is positive. The true value is positive, and the predicted value is negative; The accuracy rate is: , in, For accuracy, The actual value is negative, and the predicted value is positive. The recall rate is: , in, For recall rate, The actual value is negative, and the predicted value is negative.

[0012] In one possible implementation, the quartz sand quality testing and evaluation model includes a multimodal fusion module, a dynamic optimization decision-making module, an anomaly detection module, and an interpretable report generation module; the step of inputting the standard sample data corresponding to the quartz sand distribution characteristic data and the preliminary parameter data of the quartz sand into the trained quartz sand quality testing and evaluation model to obtain the high-purity quartz sand quality evaluation results for each purification production stage includes: The quartz sand distribution characteristic data and the standard sample data corresponding to the preliminary parameter data of the quartz sand are fused by the multimodal fusion module to obtain fused features; The fused features are processed sequentially through a dynamic optimization decision module and an anomaly detection module to obtain a quality assessment score matrix and an early warning code; The quality assessment score matrix and early warning codes are parsed by the interpretable report generation module to obtain the quality assessment results of high-purity quartz sand for each purification production stage.

[0013] Secondly, the present invention also provides an online quality detection device for high-purity quartz sand, comprising: The data acquisition module is used to acquire image data of quartz sand in each purification and production stage; The image feature extraction module is used to perform segmentation and standardization processing on the image data based on the trained image data segmentation and standardization model, to obtain the quartz sand distribution feature data corresponding to each purification and production stage, and to merge the quartz sand distribution feature data with the preset production conditions to obtain the preliminary parameter data of quartz sand for each purification and production stage. The matching module is used to match the preliminary parameter data of quartz sand in each purification and production stage with the standard sample dataset based on the trained matching model, and to determine the standard sample data corresponding to the preliminary parameter data of quartz sand. The quality assessment module is used to input the standard sample data corresponding to the quartz sand distribution characteristic data and the preliminary parameter data of the quartz sand into the trained quartz sand quality detection and assessment model to obtain the high-purity quartz sand quality assessment results for each purification and production stage.

[0014] Thirdly, the present invention also provides a quartz sand testing device, comprising: a processor and a memory; The memory stores a computer-readable program that can be executed by the processor; When the processor executes the computer-readable program, it implements the steps in the online method for detecting the quality of high-purity quartz sand as described above.

[0015] Fourthly, the present invention also provides a computer-readable storage medium for storing a computer-readable program or instructions, which, when executed by a processor, can implement the steps in the online quality detection method for high-purity quartz sand described in any one of the above-mentioned method items.

[0016] The beneficial effects of this invention are as follows: Image data is segmented and standardized based on a trained image data segmentation standardization model to obtain the distribution characteristics of quartz sand in each purification production stage. This quartz sand distribution characteristics data is then merged with preset production conditions to obtain preliminary parameter data for each purification production stage. This provides a wider range of high-purity quartz sand quality index data. Based on a trained matching model, the preliminary parameter data of quartz sand in each purification production stage is matched with a standard sample dataset to determine the standard sample data corresponding to the preliminary parameter data. By matching with the standard sample data and simultaneously detecting the quartz sand quality index data in each purification production stage, the quality of the quartz sand is guaranteed. The quartz sand distribution characteristics data and the standard sample data corresponding to the preliminary parameter data are input into a trained quartz sand quality detection and evaluation model to obtain the high-purity quartz sand quality evaluation results for each purification production stage. By evaluating the high-purity quartz sand quality in each production stage, the accuracy of quartz sand quality detection is improved. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 A flowchart illustrating an embodiment of the online quality detection method for high-purity quartz sand provided by the present invention; Figure 2 A schematic diagram of an embodiment of the online quality detection device for high-purity quartz sand provided by the present invention; Figure 3 This is a schematic diagram of an embodiment of the quartz sand testing equipment provided by the present invention. Detailed Implementation

[0019] The preferred embodiments of the present invention will be described in detail below in conjunction with the accompanying drawings, wherein the accompanying drawings constitute a part of this application and are used together with the embodiments of the present invention to illustrate the principles of the present invention, and are not used to limit the scope of the present invention.

[0020] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0021] Before demonstrating the embodiments, the following terms will be explained.

[0022] The Feature Matching Module (FMM) is a typically independently designed or packaged algorithm unit. Its core function is to find the correspondence between two or more images (or image sequences) that correspond to the same physical scene points (feature points).

[0023] This invention discloses a method, apparatus, equipment, and medium for online quality testing of high-purity quartz sand, which can be used in a computer. The method, apparatus, or computer-readable storage medium involved in this invention can be integrated with the aforementioned equipment or can be relatively independent.

[0024] One specific embodiment of the present invention discloses an online method for detecting the quality of high-purity quartz sand, which can be executed by a computer, specifically by one or more processors of the computer. Figure 1As shown, the online quality testing method for high-purity quartz sand includes: S101. Obtain image data of quartz sand in each purification production stage; It should be noted that image data of quartz sand from each purification stage of the production line is collected using automated sand particle collection equipment, automated sand particle sample preparation equipment, and automated microscopic imaging equipment. This image data includes quartz sand particle size, particle shape, inclusions, and contaminant images.

[0025] S102. Based on the trained image data segmentation and standardization model, the image data is segmented and standardized to obtain the quartz sand distribution characteristic data of each purification production stage. The quartz sand distribution characteristic data is then merged with the preset production conditions to obtain the preliminary parameter data of quartz sand for each purification production stage. It should be noted that the data on the distribution characteristics of quartz sand includes the particle size distribution characteristics, particle shape distribution characteristics, inclusion content characteristics, and contaminant content characteristics. S103. Based on the trained matching model, the preliminary parameter data of quartz sand in each purification and production stage are matched with the standard sample dataset to determine the standard sample data corresponding to the preliminary parameter data of quartz sand.

[0026] It should be noted that the matching model includes an FMM module, which uses the cosine similarity algorithm of the FMM module for identification and matching.

[0027] S104. Input the standard sample data corresponding to the distribution characteristics data of quartz sand and the preliminary parameter data of quartz sand into the trained quartz sand quality detection and evaluation model to obtain the high-purity quartz sand quality evaluation results for each purification and production stage. It should be noted that the quartz sand quality inspection and evaluation model includes a multimodal fusion module, a dynamic optimization decision-making module, an anomaly detection module, and an interpretable report generation module. The multimodal fusion module uses the Transformer cross-attention mechanism to map image features and process parameters to a unified semantic space, solving the problem of heterogeneous data alignment. The dynamic optimization decision-making module is based on the reinforcement learning PPO algorithm, aiming to minimize feature errors and improve production capacity, and dynamically generates process parameter adjustment strategies. The anomaly detection module uses an autoencoder to reconstruct errors to determine sudden anomalies and combines moving average thresholds to avoid false alarms. The interpretable report generation module integrates Gradient Class Activation Mapping (Grad-CAM) and a rule engine to generate a visualized quality heatmap and natural language optimization suggestions.

[0028] In some embodiments, in step S101, image data of quartz sand in each purification production stage is acquired. The purification production stages include physical beneficiation and chemical beneficiation. Physical beneficiation includes grinding, magnetic separation, and flotation; chemical beneficiation includes hot-pressing acid leaching and chlorination roasting. The image data of the quartz sand includes image data of quartz sand particle size, particle shape, inclusions, and contaminants. Depending on the different production stages of high-purity quartz sand, i.e., flotation, acid leaching, drying, inspection sieving, magnetic separation, and chlorination roasting, automated machinery is pre-positioned on the production line. The pre-positioned automated machinery acquires quartz sand particles from the corresponding stage. After reagent treatment, a polarizing microscope is used for automatic focusing and imaging. After each image is captured, the device automatically moves one position. The system automatically focuses and takes the next image to obtain quartz sand image data. The automated machinery includes at least an automated sand particle collection device, an automated sand particle sample preparation device, and an automated microscopic imaging device. The automated sand particle collection device collects quartz sand particles from the corresponding production line terminal. The automated sand particle sample preparation device prepares samples for quality index testing. The automated microscopic imaging device acquires relevant image data of the quartz sand particles. The reagent treatment is oil immersion, with the oil refractive index similar to that of the quartz sand particles. The quartz sand image data acquired by the automated machinery is processed and stored for subsequent data analysis and application. Online monitoring and detection by the automated machinery reduces the influence of human factors and provides a wider range of high-purity quartz sand quality index data.

[0029] In some embodiments, in step S102, the image data is segmented and standardized based on the trained image data segmentation and standardization model to obtain the quartz sand distribution feature data of each purification production stage. The quartz sand distribution feature data is then merged with preset production conditions to obtain preliminary parameter data for each purification production stage. The image segmentation and standardization model is pre-trained using a large-scale dataset, such as ImageNet. A transfer learning strategy is employed to segment and standardize the image data based on the trained model. This model includes a depthwise separable convolutional module, an inverse residual module, a linear bottleneck module, a CA attention module, and an activation function. The CA attention module is known as the Coordinate module. The Attention module uses h-swish(x) as its activation function. The segmentation standardization process involves: preprocessing the image data using a bilateral filter for noise reduction. To remove noise caused by photoelectric conversion during industrial camera image acquisition, a bilateral filtering method is used before segmentation. This method involves inputting the image data into a pre-installed bilateral filter on a central processing server, and then combining the weighted values ​​of the neighboring pixels to obtain the output pixels. The preprocessed image data is then input into a trained image segmentation standardization model. A depthwise separable convolution module extracts features from the preprocessed image data to obtain a feature map. Finally, an inverse residual module is used to... The linear bottleneck module performs linear dimensionality reduction on the feature map to obtain a low-dimensional feature map. After global average pooling of the low-dimensional feature map, the CA attention module decomposes the pooled feature map to obtain the quartz sand distribution feature data corresponding to each purification production stage. The quartz sand distribution feature data is then merged with the preset production conditions using an activation function to obtain the preliminary quartz sand parameter data for each purification production stage. This preliminary quartz sand parameter data is an intermediate feature subset containing horizontal and vertical information. Finally, the output pixels of the test series, i.e., the preprocessed image data, are subjected to size normalization and scaled to a specific size. (Height and width are 256, 4 channels correspond to 4 different feature data) After standard resolution, the input image data is segmented and standardized. After feature extraction, and a series of optimization operations using a fusion CA structure, four types of feature parameter data are finally obtained from the 4 channels: quartz sand particle size distribution feature, quartz sand particle shape distribution feature, quartz sand inclusion content feature, and quartz sand contaminant content feature. These 4 channels are 4 channels pre-embedded in the CA attention module, corresponding to the four different feature parameter data, namely, quartz sand particle size distribution feature channel, quartz sand particle shape distribution feature channel, quartz sand inclusion content channel, and quartz sand contaminant content channel. The feature channel and the quartz sand pollutant content feature channel are optimized by integrating the CA structure. The global image data information is compressed into the image data segmentation standardization model through global average pooling. Then, the global pooling is decomposed by the CA attention module and converted into a one-to-one one-dimensional feature encoding form, which is decomposed into four channels. The four types of feature parameter data output from the four channels are merged with the preset production conditions input in advance in the h-swish(x) function to form an intermediate feature subset containing horizontal and vertical information. The intermediate feature subset is the preliminary parameter data related to quartz sand particles in each purification production stage.

[0030] In some embodiments, in step S103, the preliminary parameter data of quartz sand in each purification production stage and the standard sample dataset are matched based on the trained matching model to determine the standard sample data corresponding to the preliminary parameter data of quartz sand. The trained matching model is obtained by training the matching model on the large-scale dataset ImageNet. The standard sample dataset consists of standard sample related data for each purification production stage of high-purity quartz sand. The standard sample dataset includes image-type and text-type quartz sand related parameter data. The extraction method of the standard sample dataset is as follows: collecting stable production process data of raw sand on the production line; collecting standard data of quartz sand particle size, particle shape, inclusions, and contaminants in each purification stage of raw sand; the raw sand has been verified by the qualification rate of target glass products, so its particle size, particle shape, inclusions, and contaminant related parameter data have standardity; using deep learning technology to analyze the above-mentioned raw sand particle size, particle shape, inclusions, and contaminants. The standard data of the quartz sand is processed for semantic understanding and parameter-level recognition to obtain image-type and text-type quartz sand related parameter data. The matching model includes an FMM module, and its matching process is as follows: the preliminary parameter data of quartz sand and the standard sample dataset are normalized, that is, the input features and standard sample features are L2 normalized. The cosine similarity algorithm of the FMM module is used to calculate the similarity between the normalized preliminary parameter data of quartz sand and the standard sample dataset to generate a similarity score matrix. The dot product of the normalized features is calculated by matrix multiplication to generate a similarity score matrix. The similarity score matrix is ​​judged based on a preset threshold to determine the matching result. The matching result is determined by dynamic threshold decision based on the preset thresholds of different production stages. The standard sample data corresponding to the preliminary parameter data of quartz sand is determined based on the matching result. The matching result is evaluated by accuracy, precision, and recall. The accuracy is: , in, For accuracy, The actual value is positive, and the predicted value is positive. The true value is positive, and the predicted value is negative; The accuracy rate is: , in, For accuracy, The actual value is negative, and the predicted value is positive. The recall rate was: , in, For recall rate, The actual value is negative, and the predicted value is negative.

[0031] The matching process utilizes the transfer learning technology built into the matching model to compare intermediate feature subsets containing horizontal and vertical information with standard sample data from ImageNet using the cosine similarity algorithm of the FMM module. To ensure the matching model can better learn and generalize to different feature data and process data during the matching process, the matching recognition results of the intermediate feature subsets are divided into training set, validation set, and output set in an 8:1:1 ratio. The validation set is used to evaluate the recognition-matching process using accuracy and other metrics to improve accuracy and computational efficiency. The output set contains the standard sample data corresponding to each intermediate feature subset obtained through matching. This data includes image-type and text-type quartz sand related parameter data, effectively detecting and evaluating sand sample quality indicators simultaneously under different production conditions. This improves production efficiency while ensuring product quality, and also significantly reduces labor costs, making online detection of high-purity quartz quality more universal.

[0032] In some embodiments, in step S104, the standard sample data corresponding to the quartz sand distribution characteristic data and the preliminary parameter data of quartz sand are input into the trained quartz sand quality detection and evaluation model to obtain the high-purity quartz sand quality evaluation results for each purification and production stage. The trained quartz sand quality detection and evaluation model is obtained through pre-training. The training of the quartz sand quality detection and evaluation model includes preset image recognition technology and optimization algorithms to iteratively optimize the image and text quartz sand related parameter data multiple times to minimize the error value, and finally obtain the trained quartz sand quality detection and evaluation model. The quartz sand quality detection and evaluation model includes a multimodal fusion module, a dynamic optimization decision module, an anomaly detection module, and an interpretable report generation module. The multimodal fusion module fuses the standard sample data corresponding to the quartz sand distribution characteristic data and the preliminary parameter data of quartz sand to obtain fused features. The dynamic optimization decision module and the anomaly detection module process the fused features in sequence to obtain the quality evaluation score matrix and the warning code. The interpretable report generation module processes the quality evaluation score matrix. The system analyzes the array and early warning codes to obtain the quality assessment results of high-purity quartz sand in each purification production stage. It extracts four-dimensional feature subsets (particle size / particle shape / inclusions / contaminants) and corresponding process parameter vectors (physical beneficiation conditions such as grinding, magnetic separation, and flotation, as well as various chemical beneficiation conditions such as hot-pressing acid leaching and chlorination roasting) from each purification production stage. Through a multimodal fusion module, the feature subsets and process parameters are mapped to a unified semantic space to obtain fused feature results. These fused feature results are then input into the trained quartz sand quality detection and assessment model. The model then passes through a dynamic optimization decision module and an anomaly detection module to obtain the quality assessment score matrix and early warning codes. Finally, through an interpretable report generation module, the quality assessment results are mapped into natural language process suggestions. This optimizes the evaluation results of the relevant data for each high-purity quartz sand purification production stage, resulting in a data report for the quartz sand samples corresponding to each purification stage. The high-purity quartz sand data report comprehensively considers the particle size, particle shape, inclusions, and contaminant quality indicators of each production line segment, making the report data more comprehensive and accurate.

[0033] In summary, the online quality detection method for high-purity quartz sand provided by this invention acquires image data of quartz sand in each purification production stage; performs segmentation and standardization processing on the image data based on a trained image data segmentation and standardization model to obtain the distribution feature data of quartz sand in each purification production stage; merges the quartz sand distribution feature data with preset production conditions to obtain preliminary parameter data of quartz sand in each purification production stage; matches the preliminary parameter data of quartz sand in each purification production stage with a standard sample dataset based on a trained matching model to determine the standard sample data corresponding to the preliminary parameter data of quartz sand; inputs the quartz sand distribution feature data and the standard sample data corresponding to the preliminary parameter data of quartz sand into a trained quartz sand quality detection and evaluation model to obtain the quality evaluation results of high-purity quartz sand in each purification production stage, thereby improving the accuracy of quartz sand quality detection.

[0034] To better implement the online quality detection method for high-purity quartz sand in the embodiments of the present invention, based on the online quality detection method for high-purity quartz sand, the corresponding method is as follows: Figure 2 As shown, this embodiment of the invention also provides an online quality detection device for high-purity quartz sand. The online quality detection device 200 for high-purity quartz sand includes: Data acquisition module 201 is used to acquire image data of quartz sand in each purification production stage; The image feature extraction module 202 is used to perform segmentation and standardization processing on image data based on the trained image data segmentation and standardization model, obtain the quartz sand distribution feature data corresponding to each purification and production stage, and merge the quartz sand distribution feature data with the preset production conditions to obtain the preliminary parameter data of quartz sand for each purification and production stage. The matching module 203 is used to match the preliminary parameter data of quartz sand in each purification production stage with the standard sample dataset based on the trained matching model, and to determine the standard sample data corresponding to the preliminary parameter data of quartz sand. The quality assessment module 204 is used to input the standard sample data corresponding to the distribution characteristics data and preliminary parameter data of quartz sand into the trained quartz sand quality detection and assessment model to obtain the quality assessment results of high-purity quartz sand for each purification and production stage.

[0035] like Figure 3 As shown, the present invention also provides a quartz sand testing device 300, which can be a mobile terminal, desktop computer, laptop, handheld computer, server, or other computing device. The quartz sand testing device 300 includes a processor 301, a memory 302, and a display 303. Figure 3Only some components of the quartz sand testing equipment 300 are shown; however, it should be understood that implementation of all shown components is not required, and more or fewer components may be implemented instead.

[0036] In some embodiments, the memory 302 can be an internal storage unit of the quartz sand testing device 300, such as a hard disk or memory of the quartz sand testing device 300. In other embodiments, the memory 302 can also be an external storage device of the quartz sand testing device 300, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc., equipped on the quartz sand testing device 300. Further, the memory 302 can include both internal and external storage units of the quartz sand testing device 300. The memory 302 is used to store application software and various types of data installed on the quartz sand testing device 300, such as the program code for installing the quartz sand testing device 300. The memory 302 can also be used to temporarily store data that has been output or will be output. In one embodiment, the memory 302 stores a high-purity quartz sand quality online detection program, which can be executed by the processor 301 to implement the high-purity quartz sand quality online detection method of the various embodiments of the present invention.

[0037] In some embodiments, processor 301 may be a central processing unit (CPU), microprocessor, or other data processing chip, used to run program code stored in memory 302 or process data, such as an online detection method for high-purity quartz sand quality.

[0038] In some embodiments, display 303 may be an LED display, a liquid crystal display, a touch-screen liquid crystal display, or an OLED (Organic Light-Emitting Diode) touchscreen. Display 303 is used to display identification information from the online quality testing program for high-purity quartz sand and to display a visual user interface. Components 301-303 of the quartz sand testing equipment 300 communicate with each other via a system bus.

[0039] In some embodiments, when the processor 301 executes the online detection program for high-purity quartz sand quality in the memory 302, it implements each step of the online detection method for high-purity quartz sand quality as described in the above embodiments. Since the online detection method for high-purity quartz sand quality has been described in detail above, it will not be repeated here.

[0040] Accordingly, the present invention also provides a computer-readable storage medium for storing a computer-readable program or instruction, which, when executed by a processor, can implement the steps or functions of the online detection method for high-purity quartz sand quality provided in the above-described method embodiments.

[0041] Those skilled in the art will understand that all or part of the processes of the methods described in the above embodiments can be implemented by a computer program instructing related hardware, and the program can be stored in a computer-readable storage medium. The computer-readable storage medium may be a disk, optical disk, read-only memory, or random access memory, etc.

[0042] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for online quality detection of high-purity quartz sand, characterized in that, include: Acquire image data of quartz sand in each purification and production stage; The image data is segmented and standardized based on the trained image data segmentation and standardization model to obtain the quartz sand distribution feature data of each purification and production stage. The quartz sand distribution feature data is then combined with the preset production conditions to obtain the preliminary parameter data of quartz sand for each purification and production stage. The preliminary parameter data of the quartz sand and the standard sample dataset are matched based on the trained matching model to determine the standard sample data corresponding to the preliminary parameter data of the quartz sand. The standard sample data corresponding to the quartz sand distribution characteristics data and the preliminary parameter data of the quartz sand are input into the trained quartz sand quality detection and evaluation model to obtain the high-purity quartz sand quality evaluation results for each purification and production stage.

2. The method for online quality detection of high-purity quartz sand according to claim 1, characterized in that, The purification and production process includes physical beneficiation and chemical beneficiation. Physical beneficiation includes grinding, magnetic separation and flotation. Chemical beneficiation includes hot pressing acid leaching and chlorination roasting. The image data of the quartz sand includes image data of quartz sand particle size, particle shape, inclusions and contaminants.

3. The method for online quality detection of high-purity quartz sand according to claim 1, characterized in that, The image data segmentation and standardization model includes a depthwise separable convolution module, an inverse residual module, a linear bottleneck module, a CA attention module, and an activation function; the preset production conditions include physical beneficiation and chemical beneficiation conditions; the image data is segmented and standardized based on the trained image data segmentation and standardization model to obtain quartz sand distribution feature data corresponding to each purification production stage, and the quartz sand distribution feature data is merged with the preset production conditions to obtain preliminary parameter data of quartz sand for each purification production stage, including: The image data is preprocessed; The preprocessed image data is input into a trained image data segmentation and standardization model, and the preprocessed image data is extracted using a depthwise separable convolution module to obtain a feature map. The feature map is linearly reduced in dimension by using the inverse residual module and the linear bottleneck module to obtain a low-dimensional feature map. After performing global average pooling on the low-dimensional feature map, the pooled feature map is decomposed through the CA attention module to obtain the quartz sand distribution feature data corresponding to each purification and production stage. The quartz sand distribution characteristic data is combined with preset production conditions by using an activation function to obtain preliminary parameter data of quartz sand for each purification production stage.

4. The method for online quality detection of high-purity quartz sand according to claim 3, characterized in that, The quartz sand distribution characteristics data include quartz sand particle size distribution characteristics, quartz sand particle shape distribution characteristics, quartz sand inclusion content characteristics, and quartz sand contaminant content characteristics.

5. The online quality detection method for high-purity quartz sand according to claim 3, characterized in that, The matching model includes a feature fusion module; the matching of preliminary parameter data of quartz sand from each purification production stage and standard sample datasets based on the trained matching model to determine the standard sample data corresponding to the preliminary parameter data of the quartz sand includes: The preliminary parameter data of the quartz sand and the standard sample dataset were normalized. The cosine similarity algorithm of the feature fusion module is used to calculate the similarity between the normalized preliminary parameter data of quartz sand and the standard sample dataset, and generate a similarity score matrix. The similarity score matrix is ​​evaluated based on a preset threshold to determine the matching result. Based on the matching results, standard sample data corresponding to the preliminary parameter data of the quartz sand are determined, wherein the matching results are evaluated by accuracy, precision, and recall.

6. The method for online quality detection of high-purity quartz sand according to claim 5, characterized in that, The accuracy rate is: ， in, For accuracy, The actual value is positive, and the predicted value is positive. The true value is positive, and the predicted value is negative; The accuracy rate is: ， in, is the accuracy, The actual value is negative, and the predicted value is positive. The recall rate is: ， in, For recall rate, The actual value is negative, and the predicted value is negative.

7. The method for online quality detection of high-purity quartz sand according to claim 5, characterized in that, The quartz sand quality testing and evaluation model includes a multimodal fusion module, a dynamic optimization decision-making module, an anomaly detection module, and an interpretable report generation module. The standard sample data corresponding to the quartz sand distribution characteristic data and the preliminary parameter data of the quartz sand are input into the trained quartz sand quality testing and evaluation model to obtain the high-purity quartz sand quality evaluation results for each purification production stage, including: The quartz sand distribution characteristic data and the standard sample data corresponding to the preliminary parameter data of the quartz sand are fused by the multimodal fusion module to obtain fused features; The fused features are processed sequentially through a dynamic optimization decision module and an anomaly detection module to obtain a quality assessment score matrix and an early warning code; The quality assessment score matrix and early warning codes are parsed by the interpretable report generation module to obtain the quality assessment results of high-purity quartz sand for each purification production stage.

8. An online quality testing device for high-purity quartz sand, characterized in that, include: The data acquisition module is used to acquire image data of quartz sand in each purification and production stage; The image feature extraction module is used to perform segmentation and standardization processing on the image data based on the trained image data segmentation and standardization model, to obtain the quartz sand distribution feature data corresponding to each purification and production stage, and to merge the quartz sand distribution feature data with the preset production conditions to obtain the preliminary parameter data of quartz sand for each purification and production stage. The matching module is used to match the preliminary parameter data of quartz sand in each purification and production stage with the standard sample dataset based on the trained matching model, and to determine the standard sample data corresponding to the preliminary parameter data of quartz sand. The quality assessment module is used to input the standard sample data corresponding to the quartz sand distribution characteristic data and the preliminary parameter data of the quartz sand into the trained quartz sand quality detection and assessment model to obtain the high-purity quartz sand quality assessment results for each purification and production stage.

9. A quartz sand testing device, characterized in that, Including memory and processor; The memory stores a computer-readable program that can be executed by the processor; When the processor executes the computer-readable program, it implements the steps in the online quality detection method for high-purity quartz sand as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, Used to store computer-readable programs or instructions, which, when executed by a processor, can implement the steps in the online quality detection method for high-purity quartz sand as described in any one of claims 1-7.

Images