Fly ash microelement characteristic driven active grading intelligent judgment method and system

Abstract

The application discloses a fly ash micro-element characteristic driven active grading intelligent judgment method and system, and relates to the technical field of fly ash activity detection. The method comprises the following steps: collecting fly ash sample sets under different working conditions, detecting and analyzing the micro-element characteristics and corresponding activity grades of the fly ash sample sets, and constructing a fly ash micro-element-activity grade characteristic library; a multi-branch decoupling judgment framework is built; the multi-branch decoupling judgment framework is pre-trained to generate a fly ash activity grade judge; the activity grade of the to-be-detected fly ash micro-element detection data is optimized and judged to generate a fly ash activity grade judgment result, and the fly ash activity grade judgment result is used for performing grading collaborative control. The technical problem that the fly ash activity detection process is complex, the evaluation period is long, and the fly ash activity grade judgment efficiency is low in the prior art is solved, the fly ash activity evaluation efficiency is improved, and the technical effect of realizing rapid fly ash activity grade judgment is achieved.

Description

Technical Field

[0001] This invention relates to the field of fly ash activity detection technology, specifically to an intelligent method and system for activity classification driven by fly ash micro-element characteristics. Background Technology

[0002] Fly ash is an industrial solid waste generated during coal-fired power generation. Due to its pozzolanic activity, it can be used as an important admixture in building materials such as cement and concrete. Current technologies typically evaluate the activity of fly ash through chemical composition analysis, physical performance testing, and activity index tests to determine whether it meets the requirements for different application scenarios. However, existing testing processes usually involve multiple steps, including sample preparation, experimental testing, and data analysis, making the process complex and time-consuming. This makes it difficult to meet the rapid evaluation needs given the diverse sources and frequent batch changes of fly ash. Furthermore, existing activity evaluation methods rely heavily on personnel to make comprehensive judgments based on test results. Different testing conditions and evaluation experience can easily lead to insufficient judgment efficiency, affecting rapid decision-making in the classification, utilization, and engineering applications of fly ash. Summary of the Invention

[0003] This application provides a method and system for intelligent determination of fly ash activity level driven by micro-element characteristics, which solves the technical problem that the existing fly ash activity detection process is complicated and the evaluation cycle is long, resulting in low efficiency of fly ash activity level determination.

[0004] The first aspect of this application provides an intelligent method for determining the activity classification of fly ash driven by micro-element characteristics, the method comprising: A fly ash sample set under different operating conditions was collected, and the micro-element characteristics and corresponding activity levels of the fly ash sample set were detected and analyzed to construct a fly ash micro-element-activity level feature library. A multi-branch decoupling judgment framework was built, which consists of a micro-element analysis multi-branch layer and a feature cross-fusion layer. A local calibration mechanism was introduced to pre-train the multi-branch decoupling judgment framework based on the fly ash micro-element-activity level feature library to generate a fly ash activity level determiner. The fly ash activity level determiner was used to optimize the activity level of the fly ash micro-element detection data to be tested, generating a fly ash activity level determination result, and hierarchical collaborative management and control were implemented based on the fly ash activity level determination result.

[0005] A second aspect of this application provides an intelligent determination system for activity classification driven by the micro-element characteristics of fly ash, the system comprising: Feature library construction module: Collects fly ash sample sets under different working conditions, detects and analyzes the micro-element characteristics and corresponding activity levels of the fly ash sample sets, and constructs a fly ash micro-element-activity level feature library; Judgment framework construction module: Constructs a multi-branch decoupled judgment framework, which consists of a micro-element analysis multi-branch layer and a feature cross-fusion layer; Model pre-training module: Introduces a local calibration mechanism to pre-train the multi-branch decoupled judgment framework based on the fly ash micro-element-activity level feature library, generating a fly ash activity level determiner; Intelligent judgment module: Uses the fly ash activity level determiner to optimize the activity level of the fly ash micro-element detection data to be tested, generates fly ash activity level judgment results, and performs hierarchical collaborative management and control based on the fly ash activity level judgment results.

[0006] One or more technical solutions provided in this application have at least the following technical effects or advantages: First, fly ash sample sets under different operating conditions are collected, and the micro-element characteristics and corresponding activity levels of the fly ash sample sets are detected and analyzed to construct a fly ash micro-element-activity level feature library. Next, a multi-branch decoupling judgment framework is built, consisting of a micro-element analysis multi-branch layer and a feature cross-fusion layer. Then, a local calibration mechanism is introduced to pre-train the multi-branch decoupling judgment framework based on the fly ash micro-element-activity level feature library, generating a fly ash activity level determiner. Finally, the fly ash activity level determiner is used to optimize the activity level determination of the fly ash micro-element detection data, generating fly ash activity level determination results, and implementing hierarchical collaborative management based on these results. This solves the technical problem of low efficiency in fly ash activity level determination due to the complex fly ash activity detection process and long evaluation cycle in existing technologies, achieving the technical effect of improving the efficiency of fly ash activity evaluation and realizing rapid determination of fly ash activity levels. Attached Figure Description

[0007] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying 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.

[0008] Figure 1 A schematic diagram of the intelligent determination method for activity classification driven by micro-element characteristics of fly ash provided in the embodiments of this application; Figure 2 This is a schematic diagram illustrating the change in activity level determination error before and after local calibration, provided in an embodiment of this application. Figure 3This is a schematic diagram of the intelligent determination system for activity classification driven by the micro-element characteristics of fly ash provided in the embodiments of this application.

[0009] Figure labeling: Feature library construction module 11, decision framework construction module 12, model pre-training module 13, intelligent decision module 14. Detailed Implementation

[0010] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0011] Example 1, as Figure 1 As shown, this application provides an intelligent method for determining the activity classification of fly ash driven by micro-element characteristics, wherein the method includes: Collect fly ash sample sets under different working conditions, detect and analyze the micro-element characteristics and corresponding activity levels of the fly ash sample sets, and construct a fly ash micro-element-activity level feature library.

[0012] In this embodiment, samples are collected to address the differences in operating conditions during the actual generation of fly ash, thereby obtaining fly ash sample sets with different characteristic distributions. Specifically, preset sampling rules are established based on operating condition parameters affecting the compositional characteristics and activity levels of fly ash. These preset sampling rules include operating condition stratification rules, sample batch collection rules, and sample information recording rules. The operating condition stratification rules are used to divide the fly ash generation process into different sampling intervals according to at least one of the following operating condition parameters: coal type, boiler operating load, combustion temperature, combustion time, coal powder fineness, and dust removal operating status. The sample batch collection rules are used to collect samples within each operating condition sampling interval according to a preset sampling period or preset production batch, ensuring that each operating condition sampling interval contains multiple independent fly ash samples to avoid the impact of random fluctuations in a single batch of samples on subsequent analysis results. The sample information recording rules are used to record the corresponding operating condition parameter information, collection time information, production batch information, and sample number information while collecting fly ash samples, establishing a correspondence between fly ash samples and generation operating conditions. Furthermore, according to the preset sampling rules, fly ash samples of corresponding batches are collected under different operating conditions. The collected samples are then classified and organized to form a fly ash sample set covering different combustion states, operating loads, and production batches. Specifically, for fly ash samples under the same operating conditions, multiple batches are repeatedly collected to obtain multiple sample data, characterizing the stable variation range of fly ash composition characteristics under the same operating condition. For fly ash samples under different operating conditions, sample data with differentiated microscopic composition characteristics are obtained by expanding the coverage range of operating parameters, thus enabling the fly ash sample set to reflect the variation law of fly ash composition during actual production. By collecting fly ash sample sets under different operating conditions, a sample foundation is provided for subsequent detection and analysis of fly ash micro-element characteristics, establishing the mapping relationship between micro-element characteristics and activity levels, and constructing a fly ash micro-element-activity level characteristic library.

[0013] Furthermore, the micro-element characteristics and corresponding activity levels of the fly ash sample set are detected and analyzed to construct a fly ash micro-element-activity level feature library, including: Low-damage microelement detection is performed on the fly ash sample set to extract a fly ash microelement feature set including the proportion of active oxides, the valence state of trace elements, and the degree of inter-element bonding correlation. The fly ash sample set is tested and calibrated according to a preset activity level classification system to obtain a fly ash activity level label set. The fly ash microelement feature set is mapped and bound to the corresponding fly ash activity level label set to generate a microelement feature-activity level paired dataset. Feature dimension standardization is performed on the microelement feature-activity level paired dataset to construct a fly ash microelement-activity level feature library.

[0014] Preferably, the collected fly ash sample set undergoes low-damage microelement detection to obtain elemental distribution information within the sample while preserving the original microscopic composition of the fly ash. In the specific detection process, each fly ash sample is pretreated by drying, sieving, and homogenization. The treated sample is then loaded into a microelement detection device, and elemental response data corresponding to the sample is obtained through low-damage detection methods, including at least one of X-ray fluorescence detection, laser-induced breakdown spectroscopy, and inductively coupled plasma mass spectrometry. The compositional proportions, valence state distributions, and inter-elemental relationships of different elements in the fly ash are obtained by analyzing the elemental response data. Specifically, for active oxide elements, the mass percentage of each active oxide element in the total elements of the fly ash is calculated to obtain the active oxide proportion characteristics; for trace elements with different chemical valence states, the element occurrence state is determined through valence state analysis to obtain the trace element occurrence valence state characteristics; and for the inter-elemental relationships, the bonding correlation degree between elements is calculated based on the element coexistence ratio, spatial distribution relationship, and chemical bonding characteristics to form a fly ash microelement feature set.

[0015] Furthermore, the fly ash sample set is tested and calibrated according to a preset activity level classification system. This system is established based on the fly ash activity index test results. The activity index of the fly ash samples at 28 days of age is obtained by testing the activity index under standard curing conditions. Based on this activity index, the samples are classified into three activity levels: a high activity level is defined as an activity index greater than or equal to 85%; a medium activity level is defined as an activity index between 70% and 85%; and a low activity level is defined as an activity index less than 70%. Based on the activity index test results of each fly ash sample, a corresponding activity level is matched to obtain a fly ash activity level label set. The feature parameters of each sample in the fly ash micro-element feature set are mapped and bound one-to-one with the corresponding fly ash activity level label, and the operating condition information of the corresponding samples is associated and stored to generate a micro-element feature-activity level paired dataset. Further, feature dimension standardization processing is performed on the micro-element feature-activity level paired dataset, and the scale of micro-element feature parameters of different dimensions is unified to transform each feature parameter into a unified feature space. The standardization process is expressed as follows: ,in, Represents the characteristic parameters of the i-th infinitesimal element. This represents the mean of the corresponding feature parameter. This represents the standard deviation of the corresponding feature parameter. The standardized feature parameters are represented; based on the standardized micro-element feature data, activity level labels and sample operating condition information, the features are associated and stored to construct a fly ash micro-element-activity level feature library. The feature library can characterize the correspondence between the micro-element composition features and activity levels of fly ash under different operating conditions, providing a data foundation for subsequent multi-branch decoupling judgment framework training and intelligent judgment of fly ash activity level.

[0016] A multi-branch decoupling decision framework is constructed, which consists of a micro-element analysis multi-branch layer and a feature cross-fusion layer.

[0017] Specifically, based on the data attributes of different types of micro-element features in the fly ash micro-element-activity level feature library, the standardized micro-element feature vectors are divided into feature dimensions. Feature parameters related to the content of active oxides are assigned to the constant active element analysis branch, feature parameters related to the chemical valence state of elements are assigned to the valence state active element analysis branch, and feature parameters reflecting the correlation between trace elements are assigned to the trace active element correlation analysis branch, forming a multi-branch layer of micro-element analysis composed of multiple independent analysis channels. In each analysis branch, the corresponding dimension of micro-element features are feature-encoded, converting the input element content data, valence state distribution data, and element correlation data into corresponding branch feature parameters. The correlation between feature parameters within each branch is determined based on the feature change patterns in historical samples, and the activity response features corresponding to each branch are output. Furthermore, multiple activity response features output from the micro-element analysis multi-branch layer are input into the feature cross-fusion layer. In the feature cross-fusion layer, a feature correlation matrix between different analysis branches is established. Cross-combination analysis is performed on the constant activity element response features, valence state activity element response features, and trace element synergistic response features. The influence of different feature combinations on the change of activity level is calculated. Based on the influence degree, the features of each branch are weighted and fused to generate a comprehensive activity feature vector.

[0018] A local calibration mechanism is introduced to pre-train the multi-branch decoupling judgment framework based on the fly ash micro-element-activity level feature library, thereby generating a fly ash activity level determiner.

[0019] Furthermore, a local calibration mechanism is introduced to pre-train the multi-branch decoupling decision framework based on the fly ash micro-element-activity level feature library, generating a fly ash activity level determiner, including: Based on the fly ash micro-element-activity level feature library, the multi-branch layer of micro-element analysis is trained to obtain a single-branch set of micro-elements; the output of the single-branch set of micro-elements is fed into the feature cross-fusion layer for collaborative gain pre-training to construct a global activity level determiner; according to the local calibration mechanism, a local fly ash calibration sample set is obtained; the local fly ash calibration sample set is mapped to the global activity level determiner for offset correction to generate a fly ash activity level determiner.

[0020] Specifically, the standardized micro-element feature data in the fly ash micro-element-activity level feature library are divided according to feature attributes and input into the micro-element analysis multi-branch layer for branch dimension training. This allows different analysis branches to learn the correlation between corresponding types of micro-element features and fly ash activity levels. Specifically, the constant activity element analysis branch is used to analyze the influence of changes in active oxide content on fly ash activity, the valence state activity element analysis branch is used to analyze the influence of changes in the valence state of different elements on activity release capacity, and the trace activity element correlation analysis branch is used to analyze the influence of synergistic changes among trace elements on fly ash activity. After each analysis branch completes independent feature learning, the feature response results output by each analysis branch are input into the feature cross-fusion layer. By performing correlation analysis and weight adjustment on the response results of different micro-element features, the synergistic relationship between different types of micro-element features is learned. Based on the difference between the fused comprehensive feature response result and the corresponding activity level label, parameter iterative optimization is performed to construct a global activity level determiner that can characterize the mapping relationship between fly ash micro-element features and activity levels.

[0021] Furthermore, to reduce the impact of different coal source types, combustion conditions, boiler operating parameters, and production environment differences on the micro-element composition characteristics of fly ash, a local calibration mechanism is introduced to adaptively adjust the global activity level determiner. Specifically, a local fly ash calibration sample set is obtained based on the actual application area or target production environment. This local fly ash calibration sample set includes fly ash samples with local operating conditions and corresponding activity level calibration results. Following the construction method of the fly ash micro-element-activity level feature library, micro-element detection and analysis are performed on the local fly ash calibration sample set to extract local fly ash micro-element features. These local fly ash micro-element features are then input into the global activity level determiner to obtain the response results of the local samples in the global feature space. By comparing the difference between the local fly ash micro-element feature distribution and the reference feature distribution in the fly ash micro-element-activity level feature library, the local feature offset is determined. Based on the local feature offset, the branch weights in the micro-element analysis multi-branch layer and the fusion parameters in the feature cross-fusion layer are corrected to enable the global activity level determiner to adapt to changes in the local fly ash micro-element composition. The local feature offset represents the degree of offset of the local fly ash micro-element features relative to the reference features in the global feature library, and can be expressed as: Where ΔX represents the local fly ash characteristic offset. This represents local fly ash characteristic data. The reference feature data in the fly ash micro-element-activity level feature library is represented; the branch parameters and fusion parameters in the global activity level determiner are corrected according to the local feature offset, so that the global activity level determiner adapts to the local fly ash micro-element composition feature changes, and finally generates a fly ash activity level determiner, which is used to intelligently determine the activity level of the fly ash micro-element detection data to be tested.

[0022] Furthermore, based on the fly ash micro-element-activity level feature library, the multi-branch layer of micro-element analysis is trained dimensionally to obtain a single-analysis branch set of micro-elements, including: The specific branch dimensions of the micro-element analysis multi-branch layer include constant active element analysis branch, valence state active element analysis branch, and trace active element association analysis branch. The fly ash micro-element-activity level feature library is split and labeled according to the specific branch dimensions to generate a micro-element branch dimension sub-sample set. An internal weight calibration unit is set up to perform dimension learning training and weight calibration on the constant active element analysis branch, valence state active element analysis branch, and trace active element association analysis branch based on the micro-element branch dimension sub-sample set to obtain a micro-element single analysis branch set.

[0023] Multiple feature analysis branches are pre-set in the micro-element analysis multi-branch layer, and the corresponding branch dimensions are divided according to the physicochemical properties of the micro-element characteristics of fly ash. The micro-element analysis multi-branch layer includes a constant active element analysis branch, a valence state active element analysis branch, and a trace active element correlation analysis branch. The constant active element analysis branch is used to handle the variation characteristics of the content and proportion of the main active oxides in fly ash, the valence state active element analysis branch is used to handle the variation characteristics of different occurrence valence states of trace elements, and the trace active element correlation analysis branch is used to handle the correlation variation characteristics between low-content elements.

[0024] Specifically, the fly ash micro-element-activity level feature library is feature-splitting according to the aforementioned branch dimensions, and corresponding micro-element branch dimension sub-sample sets are generated based on different micro-element feature types. Specifically, sample data containing the proportion of active oxides is assigned to the sub-sample set corresponding to the constant active element analysis branch; sample data containing element valence state characteristics is assigned to the sub-sample set corresponding to the valence state active element analysis branch; and sample data containing inter-element correlation characteristics is assigned to the sub-sample set corresponding to the trace active element correlation analysis branch. Simultaneously, the fly ash activity level label corresponding to each sub-sample set is retained to ensure that each analysis branch can learn the correlation between micro-element features and activity levels.

[0025] Furthermore, a weight calibration unit is set within the multi-branch layer of micro-element analysis. The sub-sample sets of the micro-element branch dimensions are input into the corresponding analysis branches, and the weight calibration unit dynamically adjusts the importance of different micro-element feature parameters. During training, the weight calibration unit optimizes the weight parameters corresponding to each micro-element feature based on the matching degree between the activity feature response results output by each branch and the corresponding activity level label. This ensures that micro-element features with a significant impact on the fly ash activity level receive higher weights, while suppressing the weights of features with low correlation, thereby improving the analysis capability of each analysis branch for the corresponding micro-element dimension features.

[0026] After completing independent training and weight calibration for each branch, output models for constant active element analysis branches, valence-state active element analysis branches, and trace active element correlation analysis branches are obtained. These multiple independent analysis branches are then combined to form a micro-element single analysis branch set. This micro-element single analysis branch set is used to output the activity feature response results under different micro-element dimensions, providing an input basis for subsequent feature cross-fusion layers to perform multi-dimensional feature collaborative fusion and to construct a global activity level determiner.

[0027] Furthermore, it also includes: Calculate the activity contribution set of all weight nodes in the micro-element single-analysis branch set; mark redundant nodes to be pruned whose activity contribution set is lower than a preset contribution value, and perform batch pruning on the redundant nodes to be pruned; perform accuracy correction and fine-tuning on the pruned micro-element single-analysis branch set according to the branch accuracy target.

[0028] Specifically, for the constant active element analysis branch, valence active element analysis branch, and trace active element correlation analysis branch in the micro-element single analysis branch set, the characteristic response information, weight parameters, and output changes of each weight node within each analysis branch are obtained. The corresponding activity contribution is calculated based on the contribution of each weight node in the activity level determination process. The activity contribution of each weight node is then summarized to form an activity contribution set. This set is compared with a preset contribution threshold, and weight nodes with activity contributions lower than the threshold are selected and marked as redundant nodes to be pruned. The pruning priority is determined based on the activity contribution ranking of the redundant nodes to be pruned. Pruning is performed in batches from low to high contribution. After pruning a preset number of nodes each time, the pruned nodes are... The output results of the micro-element single-analysis branch set after pruning are detected to determine whether the pruning process has caused a deviation in the activity level determination. When it is detected that the pruning has caused the branch output results to deviate from the preset accuracy range, the current pruning ratio is adjusted or pruning is stopped to avoid the loss of effective micro-element feature information. After the batch pruning process of the redundant nodes to be pruned is completed, the training samples in the fly ash micro-element-activity level feature library are re-input into the pruned micro-element single-analysis branch set. The difference analysis of the branch output results before and after pruning is performed to determine the feature response offset caused by node removal. Based on the feature response offset, the weight parameters and feature mapping relationship corresponding to the retained weight nodes are adjusted. Precision back-adjustment is performed according to the preset branch accuracy target to re-establish a stable mapping relationship between micro-element features and activity levels in each analysis branch. Finally, the micro-element single-analysis branch set optimized by redundant nodes is obtained.

[0029] Furthermore, the output of the micro-element single-analytical branch set is fed into the feature cross-fusion layer for collaborative gain pre-training to construct a global activity level determiner, including: Based on the output of the micro-element single analytical branch set, the constant basic activity contribution score, valence state activity release correction coefficient, and trace synergistic effect adjustment value are obtained. A third-order feature fusion logic and a dynamic weight iteration mechanism are built into the feature cross-fusion layer. Specifically, the third-order feature fusion logic is a third-order fusion logic of basic score-multiplier correction-fine-tuning completion. The third-order feature fusion logic and dynamic weight iteration mechanism are introduced to perform synergistic gain pre-training on the feature cross-fusion layer based on the constant basic activity contribution score, valence state activity release correction coefficient, and trace synergistic effect adjustment value to construct a global activity level determiner.

[0030] Preferably, the characteristic response results output by each analytical branch in the micro-element single analytical branch set are obtained. Specifically, the constant active element analytical branch outputs the content of active oxides and their corresponding weighted response results; the valence state active element analytical branch outputs the activity release response results corresponding to different element valence states; and the trace active element correlation analytical branch outputs the synergistic response results corresponding to the correlation changes between trace elements. Based on the output results of the different analytical branches, the characteristic response results of each branch are normalized, and the contribution of each branch to the determination of fly ash activity level is calculated in conjunction with the corresponding branch weight parameters, thereby obtaining the constant basic activity contribution score, the valence state activity release correction coefficient, and the trace synergistic contribution score. The effect adjustment value includes the constant basic activity contribution score, which is calculated by weighting the characteristic response values ​​of each active oxide in the constant active element analysis branch with the corresponding weight parameters, and is used to characterize the contribution of the main active components to the basic activity of fly ash; the valence state activity release correction coefficient is determined by the deviation relationship between the characteristic response values ​​of different existing valence states in the valence state active element analysis branch and the basic activity contribution score, and is used to characterize the degree of enhancement or weakening of the activity release capacity by the change of element valence state; the trace synergistic effect adjustment value is calculated by the degree of correlation response between the characteristics of different trace element combinations in the trace active element correlation analysis branch, and is used to characterize the supplementary influence of the trace element synergistic effect on the activity level determination result.

[0031] Furthermore, the feature cross-fusion layer is configured with three consecutive fusion processing stages: basic score, multiplier correction, and fine-tuning completion, forming a three-order feature fusion logic of basic score-multiplier correction-fine-tuning completion. Specifically, firstly, the constant basic activity contribution score is input into the basic score calculation unit, and the contribution results of each constant activity element are accumulated according to the preset basic activity contribution weight to generate a basic activity score; then, the basic activity score is input into the multiplier correction unit, and the basic activity score is proportionally adjusted according to the valence state activity release correction coefficient. When the activity release capability corresponding to the element's valence state is enhanced, the basic score is increased; when the activity release capability is reduced, the basic score is decreased, resulting in a valence state-corrected activity score; finally, the trace synergistic effect adjustment value is input into the fine-tuning completion unit, and the valence state-corrected activity score is compensated and adjusted according to the trace element correlation response results to obtain the fused comprehensive activity feature value.

[0032] A dynamic weight iteration mechanism is introduced in the above three-stage fusion process. The fused comprehensive activity feature value is compared with the corresponding activity level label in the fly ash micro-element-activity level feature library. The fusion weights corresponding to the constant basic activity contribution score, valence state activity release correction coefficient, and trace synergistic effect adjustment value are adjusted according to the judgment deviation. Specifically, the fusion weight is increased for feature dimensions that cause large judgment deviations, and decreased for feature dimensions that have little impact on the judgment result. The fusion parameters are updated iteratively through multiple rounds of sample training, so that the feature cross-fusion layer can adaptively adjust the fusion ratio according to different micro-element combination features. After synergistic gain pre-training, a global activity level determiner is established using the trained micro-element branch weight parameters and fusion parameters. The global activity level determiner can output the corresponding activity level judgment result based on the input fly ash micro-element feature data.

[0033] Furthermore, mapping the local fly ash calibration sample set to the global activity level determiner for offset correction, and generating the fly ash activity level determiner, includes: The local fly ash calibration sample set is mapped to the high-dimensional feature space of the global activity level determiner for alignment processing to obtain the local fly ash feature sample set; the local fly ash feature offset set is calculated and quantified to compare the local fly ash feature sample set with the features of the same type of samples in the fly ash micro-element-activity level feature library; the global activity level determiner is offset corrected based on the local fly ash feature offset set to generate the fly ash activity level determiner.

[0034] Preferably, the local fly ash calibration sample set is input into the global activity level determiner. The micro-element analysis multi-branch layer in the global activity level determiner is invoked to extract features from the constant active element features, valence active element features, and trace element correlation features corresponding to the local fly ash samples. The feature response results output by each analysis branch are then transformed into a unified high-dimensional feature space according to the feature mapping relationship formed during the training phase of the global activity level determiner. Within the high-dimensional feature space, based on the activity level label corresponding to the local fly ash samples, the local fly ash feature samples and the fly ash micro-elements are compared. - Samples in the activity level feature library are classified and matched. First, reference samples with the same activity level label are screened. Then, the corresponding reference sample set is determined based on the similarity of micro-element features. The feature distribution of local fly ash samples is aligned with that of the reference sample set to obtain the local fly ash feature sample set after feature space mapping. Further, for the local fly ash feature sample set after feature alignment, the degree of difference between the local sample features and the corresponding reference sample features is calculated according to three dimensions: constant active element features, valence state active element features, and trace element correlation features. The differences between each feature dimension and the corresponding reference sample features are statistically analyzed. The deviation direction, deviation magnitude, and continuous deviation trend under different degrees are used to form a corresponding local fly ash feature offset set. Among them, for feature dimensions with a large degree of deviation, it is determined that the micro-element features corresponding to this dimension are significantly affected by changes in local coal source, production conditions, or raw material composition, and key corrections are required. For feature dimensions with a small degree of deviation, the original global judgment parameters remain unchanged. Further, the global activity level determiner is offset corrected according to the local fly ash feature offset set. Specifically, the features of the corresponding analytical branches in the multi-branch layer of micro-element analysis are adjusted according to the deviation direction of each micro-element feature dimension. Weights are assigned to each branch to adapt to changes in the micro-element composition of local fly ash. Simultaneously, the fusion weights in the feature cross-fusion layer are adjusted based on the degree of deviation in the output results of each branch, altering the proportions of constant active element contribution, valence state active release correction, and trace element synergistic regulation in the final activity level determination process. After parameter adjustment, the local fly ash calibration sample set is re-input into the corrected global activity level determiner. The output activity level result is compared with the actual calibration level. Based on the comparison result, the deviation correction parameters are further adjusted until the output result meets the preset determination requirements, ultimately generating the fly ash activity level determiner.

[0035] Furthermore, the global activity level determiner is offset-corrected based on the local fly ash feature offset set to generate a fly ash activity level determiner, including: The local fly ash feature offset set is input into the global activity level determiner in batches according to the priority from largest to smallest for directional iterative offset correction to obtain the adaptive activity level determiner; the element combination rules of the local fly ash feature offset set are statistically analyzed to construct the local fly ash element activity adaptation rules; the local fly ash element activity adaptation rules are introduced to constrain and correct the adaptive activity level determiner to generate the fly ash activity level determiner.

[0036] First, the offset degrees of each feature dimension in the local fly ash feature offset set are sorted. The correction priority is determined based on the deviation of each micro-element feature from the corresponding reference feature in the fly ash micro-element-activity level feature library. The corresponding feature offsets are then input into the global activity level determiner in batches, from largest to smallest, for targeted iterative offset correction. During each batch of feature offset correction, the corresponding micro-element analysis branch and feature fusion parameters in the global activity level determiner are located based on the micro-element feature dimension of the current batch. For constant active element features with larger offsets, the constant active element analysis branch is adjusted. The feature weights in the branch are adjusted to adjust the release correction parameters in the valence active element analysis branch for features with large offsets, and the association weights in the trace active element association analysis branch for trace element association features with large offsets. At the same time, the fusion ratio of the corresponding feature dimension in the feature cross-fusion layer is corrected. After the offset correction of the current batch is completed, the activity level is re-determined using the local fly ash calibration sample set. The correction direction of the next batch is determined based on the deviation between the determination result and the actual calibration level. The process of feature offset input, parameter adjustment and result verification is repeated until an adaptive activity level determiner that adapts to the local fly ash feature distribution is obtained.

[0037] Furthermore, statistical analysis is performed on the element combination variation patterns within the local fly ash characteristic offset set to obtain the correspondence between different micro-element combinations and activity level changes. Specifically, based on the changes in active oxide content, element valence state combinations, and trace element correlations in local fly ash samples, cluster analysis is conducted on samples with the same or similar activity levels. The distribution of activity levels under different element combination conditions is statistically analyzed, and element combination features with stable influence on activity level changes are screened to form local fly ash element activity adaptation rules that include element combination relationships, activity response directions, and applicable activity level ranges.

[0038] Finally, the local fly ash elemental activity adaptation rules are introduced into the adaptation activity level determiner for constraint correction. Specifically, after the adaptation activity level determiner outputs the activity level determination result, the micro-element combination features corresponding to the output result are matched with the local fly ash elemental activity adaptation rules. When the determination result is found to be inconsistent with the local elemental combination rules, the micro-element analysis branch weight, feature fusion weight, and level output threshold are corrected according to the elemental activity adaptation rules to make the determination result conform to the local fly ash elemental composition rules. After rule constraint correction, a fly ash activity level determiner is generated, which can adapt to the changes in local fly ash micro-element characteristics while maintaining global determination capabilities, thereby improving the accuracy and stability of activity level determination for fly ash from different sources.

[0039] like Figure 2 As shown, without local calibration, the global activity level determiner has a certain judgment bias due to the differences in micro-element composition of fly ash samples from different sources. After correcting the feature shift of the global activity level determiner by using a local fly ash calibration sample set, the activity level determination error of each batch of samples was reduced, indicating that the local calibration mechanism can effectively improve the adaptability of the determiner to fly ash samples from different sources.

[0040] The fly ash activity level determiner is used to optimize and determine the activity level of the fly ash micro-element detection data, generate fly ash activity level determination results, and implement hierarchical collaborative management and control based on the fly ash activity level determination results.

[0041] Furthermore, the fly ash activity level determiner is used to optimize and determine the activity level of the fly ash micro-element detection data to be tested, generating a fly ash activity level determination result, including: The accumulated real activity data of the samples to be tested are sent back to the fly ash activity level determiner for deviation self-correction, resulting in an optimized activity level determiner. The optimized activity level determiner is then used to optimize the activity level of the fly ash micro-element detection data, generating a fly ash activity level determination result.

[0042] During the operation of the fly ash activity level determiner, real activity data of the fly ash samples under test are collected in actual application. This real activity data is then transmitted back to the fly ash activity level determiner for deviation self-correction. Specifically, after obtaining the fly ash samples to be tested, some samples undergo actual activity testing according to the standard activity testing procedure to obtain the real activity index and activity level label corresponding to the samples. Simultaneously, the micro-element detection data of the corresponding samples is input into the fly ash activity level determiner to obtain the initial activity level prediction result output by the fly ash activity level determiner. The real activity level label is then compared with the initial activity level prediction result to determine the different micro-elements. The judgment deviation corresponding to the elemental analysis branch is analyzed. Based on this judgment deviation, the parameters of the fly ash activity level determiner are corrected. The deviation between the actual activity level and the predicted activity level is back-mapped to the micro-elemental analysis multi-branch layer and the feature cross-fusion layer. When there is a deviation between the output result of the constant active element analysis branch and the actual activity level, the branch weights corresponding to the active oxide features are adjusted. When there is a shift in the output result of the valence state active element analysis branch, the mapping parameters of the corresponding element's valence state features are adjusted. When there is a deviation in the output result of the trace active element correlation analysis branch, the weight parameters corresponding to the trace element correlation features are recalculated. Simultaneously, the comprehensive analysis of the output results of multiple analysis branches and the actual activity level is performed. To address the deviation, the fusion ratio in the feature cross-fusion layer is adjusted to match the contribution relationship of different dimensions of micro-element features in the activity level determination process with the actual change law of fly ash activity. After parameter adjustment, the output result of the updated fly ash activity level determiner is compared with the actual activity level again. When the updated determination deviation is less than the preset correction threshold, parameter adjustment is stopped, and an optimized activity level determiner with self-correction of deviation is obtained. Furthermore, the optimized activity level determiner is used to optimize the activity level determination of the micro-element detection data of the fly ash to be tested. Micro-element detection is performed on the fly ash to be tested to obtain the proportion of active oxides, the valence state of trace elements, and the bonding correlation between elements corresponding to the fly ash to be tested. The data is standardized and transformed according to the feature processing method in the fly ash micro-element-activity level feature library to form the fly ash micro-element feature data to be tested. The fly ash micro-element feature data to be tested is input into the activity level optimization and determination device. The micro-element features of the corresponding dimensions are independently analyzed through the constant active element analysis branch, the valence state active element analysis branch, and the trace active element association analysis branch to obtain the activity feature response results of different dimensions. Then, the activity feature response results are input into the feature cross-fusion layer. The comprehensive activity feature representation is obtained by calculating the fusion weight. The comprehensive activity feature representation is matched with the preset activity level classification rules to output the activity level determination result of the fly ash to be tested.

[0043] The hierarchical collaborative management based on the fly ash activity level determination results includes: obtaining the activity level category and corresponding activity evaluation parameters from the fly ash activity level determination results; matching the activity level category with preset hierarchical management rules to determine the corresponding fly ash application level and allocation priority; wherein, the preset hierarchical management rules are established based on the fly ash activity index, active oxide content, and application requirements, and are used to determine the treatment methods and application parameters corresponding to different activity levels of fly ash. In specific implementation, a three-level activity level classification rule is established based on the fly ash activity index test results: when the fly ash activity index is not lower than 85%, the corresponding fly ash is classified as high-activity level fly ash; when the fly ash activity index is in the range of 70% to 85%, the corresponding fly ash is classified as medium-activity level fly ash; when the fly ash activity index is lower than 70%, the corresponding fly ash is classified as low-activity level fly ash. Meanwhile, corresponding blending parameters are set according to the application requirements of different activity levels. The target blending ratio range for high-activity fly ash is set to 20% to 40%, and it is given priority in the preparation of cementitious materials with high activity requirements. The target blending ratio range for medium-activity fly ash is set to 10% to 25%, and it is allocated to the application of ordinary concrete. The target blending ratio range for low-activity fly ash is set to no more than 10%, and it is used or used as an auxiliary blend after pretreatment according to the actual activity index.

[0044] Specifically, based on the fly ash activity level determination results, batches of fly ash to be treated are labeled with their activity levels. Fly ash samples with the same or similar activity levels are grouped into the same control unit, and a target utilization strategy is determined according to the activity level corresponding to each control unit. When the determination result corresponds to a high activity level, the batch of fly ash is marked as a high activity resource unit, and priority application blending parameters are generated, allocating it to the preparation stage of cementitious materials with high activity requirements. When the determination result corresponds to a medium activity level, the corresponding blending ratio is calculated based on the activity index range, and conventional application blending parameters are generated, allocating it to the ordinary building material blending application stage. When the determination result corresponds to a low activity level, downgrade utilization or pretreatment parameters are generated based on the degree of insufficient activity, and modification treatment, auxiliary blending, or resource utilization control are implemented.

[0045] In summary, the embodiments of this application have at least the following technical effects: First, fly ash sample sets under different operating conditions are collected, and the micro-element characteristics and corresponding activity levels of the fly ash sample sets are detected and analyzed to construct a fly ash micro-element-activity level feature library. Next, a multi-branch decoupling judgment framework is built, consisting of a micro-element analysis multi-branch layer and a feature cross-fusion layer. Then, a local calibration mechanism is introduced to pre-train the multi-branch decoupling judgment framework based on the fly ash micro-element-activity level feature library, generating a fly ash activity level determiner. Finally, the fly ash activity level determiner is used to optimize the activity level determination of the fly ash micro-element detection data, generating fly ash activity level determination results, and implementing hierarchical collaborative management based on these results. This solves the technical problem of low efficiency in fly ash activity level determination due to the complex fly ash activity detection process and long evaluation cycle in existing technologies, achieving the technical effect of improving the efficiency of fly ash activity evaluation and realizing rapid determination of fly ash activity levels.

[0046] Example 2, based on the same inventive concept as the fly ash micro-element characteristic-driven intelligent determination method for activity classification in the previous examples, such as... Figure 3 As shown, this application provides an intelligent determination system for activity classification driven by the micro-element characteristics of fly ash, wherein the system includes: Feature library construction module 11: Collects fly ash sample sets under different working conditions, detects and analyzes the micro-element characteristics and corresponding activity levels of the fly ash sample sets, and constructs a fly ash micro-element-activity level feature library; Judgment framework construction module 12: Constructs a multi-branch decoupled judgment framework, which consists of a micro-element analysis multi-branch layer and a feature cross-fusion layer; Model pre-training module 13: Introduces a local calibration mechanism to pre-train the multi-branch decoupled judgment framework based on the fly ash micro-element-activity level feature library, generating a fly ash activity level determiner; Intelligent judgment module 14: Uses the fly ash activity level determiner to optimize the activity level of the fly ash micro-element detection data to be tested, generates a fly ash activity level determination result, and performs hierarchical collaborative management and control through the fly ash activity level determination result.

[0047] Furthermore, the feature library construction module 11 is used to perform the following methods: Low-damage microelement detection is performed on the fly ash sample set to extract a fly ash microelement feature set including the proportion of active oxides, the valence state of trace elements, and the degree of inter-element bonding correlation. The fly ash sample set is tested and calibrated according to a preset activity level classification system to obtain a fly ash activity level label set. The fly ash microelement feature set is mapped and bound to the corresponding fly ash activity level label set to generate a microelement feature-activity level paired dataset. Feature dimension standardization is performed on the microelement feature-activity level paired dataset to construct a fly ash microelement-activity level feature library.

[0048] Furthermore, the model pre-training module 13 is used to perform the following methods: Based on the fly ash micro-element-activity level feature library, the multi-branch layer of micro-element analysis is trained to obtain a single-branch set of micro-elements; the output of the single-branch set of micro-elements is fed into the feature cross-fusion layer for collaborative gain pre-training to construct a global activity level determiner; according to the local calibration mechanism, a local fly ash calibration sample set is obtained; the local fly ash calibration sample set is mapped to the global activity level determiner for offset correction to generate a fly ash activity level determiner.

[0049] Furthermore, the model pre-training module 13 is used to perform the following methods: The specific branch dimensions of the micro-element analysis multi-branch layer include constant active element analysis branch, valence state active element analysis branch, and trace active element association analysis branch. The fly ash micro-element-activity level feature library is split and labeled according to the specific branch dimensions to generate a micro-element branch dimension sub-sample set. An internal weight calibration unit is set up to perform dimension learning training and weight calibration on the constant active element analysis branch, valence state active element analysis branch, and trace active element association analysis branch based on the micro-element branch dimension sub-sample set to obtain a micro-element single analysis branch set.

[0050] Furthermore, the model pre-training module 13 is used to perform the following methods: Calculate the activity contribution set of all weight nodes in the micro-element single-analysis branch set; mark redundant nodes to be pruned whose activity contribution set is lower than a preset contribution value, and perform batch pruning on the redundant nodes to be pruned; perform accuracy correction and fine-tuning on the pruned micro-element single-analysis branch set according to the branch accuracy target.

[0051] Furthermore, the model pre-training module 13 is used to perform the following methods: Based on the output of the micro-element single analytical branch set, the constant basic activity contribution score, valence state activity release correction coefficient, and trace synergistic effect adjustment value are obtained. A third-order feature fusion logic and a dynamic weight iteration mechanism are built into the feature cross-fusion layer. Specifically, the third-order feature fusion logic is a third-order fusion logic of basic score-multiplier correction-fine-tuning completion. The third-order feature fusion logic and dynamic weight iteration mechanism are introduced to perform synergistic gain pre-training on the feature cross-fusion layer based on the constant basic activity contribution score, valence state activity release correction coefficient, and trace synergistic effect adjustment value to construct a global activity level determiner.

[0052] Furthermore, the model pre-training module 13 is used to perform the following methods: The local fly ash calibration sample set is mapped to the high-dimensional feature space of the global activity level determiner for alignment processing to obtain the local fly ash feature sample set; the local fly ash feature offset set is calculated and quantified to compare the local fly ash feature sample set with the features of the same type of samples in the fly ash micro-element-activity level feature library; the global activity level determiner is offset corrected based on the local fly ash feature offset set to generate the fly ash activity level determiner.

[0053] Furthermore, the model pre-training module 13 is used to perform the following methods: The local fly ash feature offset set is input into the global activity level determiner in batches according to the priority from largest to smallest for directional iterative offset correction to obtain the adaptive activity level determiner; the element combination rules of the local fly ash feature offset set are statistically analyzed to construct the local fly ash element activity adaptation rules; the local fly ash element activity adaptation rules are introduced to constrain and correct the adaptive activity level determiner to generate the fly ash activity level determiner.

[0054] Furthermore, the intelligent determination module 14 is used to perform the following method: The accumulated real activity data of the samples to be tested are sent back to the fly ash activity level determiner for deviation self-correction, resulting in an optimized activity level determiner. The optimized activity level determiner is then used to optimize the activity level of the fly ash micro-element detection data, generating a fly ash activity level determination result.

[0055] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A method for intelligent determination of activity classification driven by the micro-element characteristics of fly ash, characterized in that, The method includes: Collect fly ash sample sets under different working conditions, detect and analyze the micro-element characteristics and corresponding activity levels of the fly ash sample sets, and construct a fly ash micro-element-activity level feature library; A multi-branch decoupling decision framework is constructed, which consists of a micro-element analysis multi-branch layer and a feature cross-fusion layer. A local calibration mechanism is introduced to pre-train the multi-branch decoupling judgment framework based on the fly ash micro-element-activity level feature library to generate a fly ash activity level determiner. The fly ash activity level determiner is used to optimize and determine the activity level of the fly ash micro-element detection data, generate fly ash activity level determination results, and implement hierarchical collaborative management and control based on the fly ash activity level determination results.

2. The intelligent determination method for activity classification driven by the micro-element characteristics of fly ash as described in claim 1, characterized in that, The micro-element characteristics and corresponding activity levels of the fly ash sample set were detected and analyzed to construct a fly ash micro-element-activity level feature library, including: Low-damage micro-element detection was performed on the fly ash sample set to extract a micro-element feature set of fly ash, including the proportion of active oxides, the valence state of trace elements, and the degree of bonding between elements. The fly ash sample set was tested and calibrated according to the preset activity level classification system to obtain a fly ash activity level label set. The fly ash micro-element feature set is mapped and bound to the corresponding fly ash activity level label set to generate a micro-element feature-activity level paired dataset. The feature dimension standardization was performed on the micro-element feature-activity level paired dataset to construct a fly ash micro-element-activity level feature library.

3. The intelligent determination method for activity classification driven by the micro-element characteristics of fly ash as described in claim 1, characterized in that, A local calibration mechanism is introduced to pre-train the multi-branch decoupling decision framework based on the fly ash micro-element-activity level feature library, generating a fly ash activity level determiner, including: Based on the fly ash micro-element-activity level feature library, the micro-element analysis multi-branch layer is trained to obtain the micro-element single-analysis branch set. The output of the micro-element single analytical branch set is fed into the feature cross-fusion layer for collaborative gain pre-training to construct a global activity level determiner. According to the local calibration mechanism, obtain a local fly ash calibration sample set; The local fly ash calibration sample set is mapped to the global activity level determiner for offset correction, thereby generating the fly ash activity level determiner.

4. The intelligent determination method for activity classification driven by the micro-element characteristics of fly ash as described in claim 3, characterized in that, Based on the fly ash micro-element-activity level feature library, the multi-branch layer of micro-element analysis is trained to obtain a single-analysis branch set of micro-elements, including: The specific branch dimensions of the micro-element analysis multi-branch layer include constant active element analysis branch, valence state active element analysis branch, and trace active element correlation analysis branch. The fly ash micro-element-activity level feature library is split and labeled according to specific branch dimensions to generate micro-element branch dimension sub-sample sets; An internal weight calibration unit performs dimension learning training and weight calibration on the constant active element analysis branch, valence active element analysis branch, and trace active element association analysis branch based on the micro-element branch dimension sub-sample set, to obtain a micro-element single analysis branch set.

5. The intelligent determination method for activity classification driven by the micro-element characteristics of fly ash as described in claim 4, characterized in that, The method further includes: Calculate the set of active contribution values ​​of all weight nodes in the single analytical branch set of the micro-element; Mark the redundant nodes to be pruned that are below the preset contribution level, and perform batch pruning on the redundant nodes to be pruned. The precision of the micro-element single-analytic branch set after pruning is fine-tuned according to the branch accuracy target.

6. The intelligent determination method for activity classification driven by the micro-element characteristics of fly ash as described in claim 3, characterized in that, The output of the micro-element single-analytical branch set is fed into the feature cross-fusion layer for collaborative gain pre-training to construct a global activity level determiner, including: Based on the output of the micro-element single analytical branch set, the constant basic activity contribution score, the valence state activity release correction coefficient, and the trace synergistic effect adjustment value are obtained. The feature cross-fusion layer incorporates a third-order feature fusion logic and a dynamic weight iteration mechanism. Specifically, the third-order feature fusion logic consists of a basic score, a multiplier correction, and a fine-tuning completion. The third-order feature fusion logic and dynamic weight iteration mechanism are introduced to perform synergistic gain pre-training on the feature cross-fusion layer based on the constant basic activity contribution score, valence state activity release correction coefficient and trace synergistic effect adjustment value, and a global activity level determiner is constructed.

7. The intelligent determination method for activity classification driven by the micro-element characteristics of fly ash as described in claim 3, characterized in that, Mapping the local fly ash calibration sample set to the global activity level determiner for offset correction generates the fly ash activity level determiner, including: The local fly ash calibration sample set is mapped to the high-dimensional feature space of the global activity level determiner for alignment processing to obtain the local fly ash feature sample set. Calculate and quantify the local fly ash feature offset set between the local fly ash feature sample set and the features of the same type of sample in the fly ash micro-element-activity level feature library; The global activity level determiner is offset-corrected based on the local fly ash feature offset set to generate a fly ash activity level determiner.

8. The intelligent determination method for activity classification driven by the micro-element characteristics of fly ash as described in claim 7, characterized in that, The global activity level determiner is offset-corrected based on the local fly ash feature offset set to generate a fly ash activity level determiner, including: The local fly ash feature offset set is input into the global activity level determiner in batches according to the priority from large to small for directional iterative offset correction, so as to obtain the adaptive activity level determiner. The element combination rules of the local fly ash feature offset set are statistically analyzed to construct the local fly ash element activity adaptation rules. The local fly ash elemental activity matching rule is introduced to constrain and correct the matching activity level determiner, thereby generating a fly ash activity level determiner.

9. The intelligent determination method for activity classification driven by the micro-element characteristics of fly ash as described in claim 1, characterized in that, The fly ash activity level determiner is used to optimize and determine the activity level of the fly ash micro-element detection data, generating fly ash activity level determination results, including: The accumulated real activity data of the samples to be tested are sent back to the fly ash activity level determiner for deviation self-correction, and an optimized activity level determiner is obtained. The activity level optimization and determination device is used to optimize and determine the activity level of the micro-element detection data of the fly ash to be tested, and the fly ash activity level determination result is generated.

10. An intelligent determination system for activity classification driven by the micro-element characteristics of fly ash, characterized in that, The system is used to implement the fly ash micro-element characteristic-driven intelligent determination method for activity classification according to any one of claims 1-9, the system comprising: Feature library construction module: Collect fly ash sample sets under different working conditions, detect and analyze the micro-element characteristics and corresponding activity levels of the fly ash sample sets, and construct a fly ash micro-element-activity level feature library; Module for constructing a decision framework: Constructing a multi-branch decoupling decision framework, which consists of a micro-element parsing multi-branch layer and a feature cross-fusion layer; Model pre-training module: Introduces a local calibration mechanism to pre-train the multi-branch decoupling judgment framework based on the fly ash micro-element-activity level feature library, and generates a fly ash activity level determiner. Intelligent Judgment Module: The fly ash activity level determiner is used to optimize and determine the activity level of the fly ash micro-element detection data, generate fly ash activity level determination results, and perform hierarchical collaborative management and control based on the fly ash activity level determination results.

Images