A dynamic prediction method for cement negative pressure screen residue

Abstract

The present application relates to the technical field of industrial data processing, in particular to a cement negative pressure screen analyzer residue amount dynamic prediction method, comprising: constructing a production stage database, collecting samples to be detected through distributed nodes, and establishing the relevance of the samples to historical screening maps by using association rule mining to generate a weighted sample data cube; analyzing multi-dimensional parameters in a dynamic screening log library through a data processing model to construct a screen state topology relationship; using a multi-dimensional index algorithm to calculate the similarity measure of sensor data flow and particle distribution database to generate data correlation factors; establishing a state discrimination rule library according to the screen state topology relationship and the data correlation factors, and performing residue amount prediction through corresponding data processing methods; and performing weight vector aggregation on the predicted value to output a residue amount prediction result set. The present application realizes high-precision residue amount prediction in complex production scenarios such as insufficient sample representation, frequent screen clogging and environmental factor disturbance.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of industrial data processing, and particularly to a dynamic prediction method for the sieve residue of a cement negative pressure sieve analyzer. BACKGROUND

[0002] At present, the prediction method for the sieve residue of a cement negative pressure sieve analyzer is generally based on a static structured data architecture, mainly by constructing an index relationship between attribute data tables (including particle size distribution, chemical composition, etc.) of limited samples and historical sieve records, and using a traditional database query method to realize prediction calculation. The existing system stores production sample data, sieve equipment operation logs and environmental sensing parameters in different database tables, only supports information retrieval under single dimension and static conditions, and is difficult to carry out dynamic analysis of data across variables and stages.

[0003] The existing method often uses a relational database to match and conditionally filter different data sources through a structured query language (SQL). Some systems realize the basic mapping between sample attributes and sieve residue results based on a static weight model, but lack the ability to deeply mine the implicit relationships between high-dimensional and multi-variable data. This isolated and static data organization mode limits the accuracy and adaptability of data prediction in complex production scenarios, especially in the case of insufficient sample representativeness, frequent sieve screen clogging and environmental factor disturbance, etc. The traditional model is difficult to ensure the stability and real-time performance of the prediction results.

[0004] Therefore, the existing technology has not established a unified multi-dimensional data fusion and dynamic modeling mechanism, and cannot realize comprehensive collection, indexing and adaptive analysis of key information in the screening process. In view of the problems of process nonlinearity, state dynamic evolution and external environmental disturbance in the cement screening process, an information processing method oriented to data structure optimization and dynamic association mining is needed to support high-precision sieve residue prediction and linkage judgment of complex screening states.

[0005] Therefore, a dynamic prediction method for the sieve residue of a cement negative pressure sieve analyzer is proposed. SUMMARY

[0006] The present application relates to the technical field of industrial data processing, and particularly to a dynamic prediction method for the sieve residue of a cement negative pressure sieve analyzer.

[0007] To achieve the above-mentioned purpose, the present application provides the following technical solutions:

[0008] A dynamic prediction method for the sieve residue of a cement negative pressure sieve analyzer, comprising:

[0009] A production stage database is constructed, and a to-be-detected sample is collected by a distributed node to obtain an attribute data table of the to-be-detected sample; an association rule mining is used to establish the association between the attribute data table and a historical screening atlas, and a weighted sample data cube is generated;

[0010] A dynamic screening log library is created according to the weighted sample data cube, and a data processing model is used to analyze multi-dimensional parameters in the dynamic screening log library to construct a screen mesh state topology relationship;

[0011] Sensing data streams are obtained by a multi-source sensor, a multi-dimensional index algorithm is used to calculate the similarity measurement between the sensing data streams and a particle distribution database, and a data correlation factor is generated;

[0012] A first prediction value is output by calling a data relationship model, and a state discrimination rule library is established according to the screen mesh state topology relationship and the data correlation factor: when a first rule is met, the first prediction value is directly output; when a second rule is met, a first compensation calibration is performed based on the screen mesh state topology relationship to generate a second prediction value; when a third rule is met, a second compensation calibration is performed by an environmental influence model to generate a third prediction value;

[0013] The first prediction value, the second prediction value and the third prediction value are aggregated by a weight vector to output a screen residue amount prediction result set.

[0014] Preferably, the generation process of the weighted sample data cube comprises:

[0015] A cement production process is divided into multiple production stages, and N distributed sampling nodes are configured for each production stage to collect to-be-detected samples;

[0016] Attribute data of each to-be-detected sample is obtained, including particle size distribution, particle morphology, particle spacing and humidity, an attribute data table is constructed, and stored in the production stage database;

[0017] An association rule mining is used to analyze the potential association between the attribute data table and a historical screening atlas, and an association rule path is established;

[0018] The importance of each to-be-detected sample to a preset screening index is calculated according to the association rule path, and a representative weight is assigned to each to-be-detected sample;

[0019] The to-be-detected samples are combined by the representative weight to generate the weighted sample data cube.

[0020] Preferably, the construction process of the screen mesh state topology relationship comprises:

[0021] The weighted sample data cube is dynamically screened to monitor multi-dimensional parameters in real time, including screen hole flow, screen mesh clogging degree and vibration frequency, and a dynamic screening log library is created.

[0022] analyzing the multi-dimensional parameters, solving the clogging rate by simultaneous equations, and identifying and predicting the clogging trend;

[0023] constructing a screen state topology relationship according to the clogging rate and the clogging trend.

[0024] Preferably, the generation process of the data correlation factor comprises:

[0025] normalizing the sensing data stream obtained through the multi-source sensor, the sensing data stream comprising humidity, temperature, air pressure and dust concentration;

[0026] constructing a multi-dimensional feature vector based on the sensing data stream, performing similarity matching on a particle distribution database by using a multi-dimensional index algorithm to obtain a similarity measure value, and extracting K reference samples with the highest similarity measure value; the particle distribution database contains a plurality of paired samples of historical environmental characteristics and corresponding particle distribution curves;

[0027] combining the similarity measure value to generate the data correlation factor reflecting the relationship between the current environmental state and the historical particle distribution.

[0028] Preferably, the state discrimination rule base specifically comprises:

[0029] when the clogging trend does not satisfy the preset clogging condition and the data correlation factor is lower than the preset correlation degree threshold, it is determined that the first rule is met;

[0030] when the clogging trend satisfies the preset clogging condition and the data correlation factor is lower than the preset correlation degree threshold, it is determined that the second rule is met;

[0031] when the clogging trend does not satisfy the preset clogging condition and the data correlation factor is higher than the preset correlation degree threshold, it is determined that the third rule is met.

[0032] Preferably, the specific process of the first compensation calibration comprises:

[0033] combining the screen state topology relationship to analyze the deviation influence of the multi-dimensional parameters on the screen residue prediction;

[0034] calculating a compensation amount for correcting the second predicted value according to the deviation influence;

[0035] combining the calculated compensation amount with the first predicted value to generate the second predicted value after the first compensation calibration.

[0036] Preferably, the specific process of the second compensation calibration comprises:

[0037] A coupling mapping relationship between environmental factors and particle motion characteristics is constructed through an environmental impact model, including a temperature-diffusion rate function, a humidity-aggregation coefficient function, an air pressure-particle migration probability function, and a dust concentration-sieve aperture permeability function;

[0038] The coupling mapping relationship is applied to a current sensor data stream to dynamically calculate an environmental interference influence coefficient matrix;

[0039] A particle distribution model is corrected according to the influence coefficient matrix to inversely calculate a sieve residue deviation trend in a sieving process;

[0040] A first predicted value is adjusted in deviation based on the inverse calculation result to generate a third predicted value that is compensated and calibrated.

[0041] Preferably, the output process of the sieve residue amount prediction result set comprises:

[0042] The first predicted value, the second predicted value and the third predicted value generated according to different rules are integrated;

[0043] Weights are assigned based on the reliability and importance of each group of predicted values, and the three groups of predicted values are weighted and fused to obtain a comprehensive sieve residue amount predicted value;

[0044] The comprehensive sieve residue amount predicted value is output in a structured format as the final sieve residue amount prediction result set.

[0045] Compared with the prior art, the present application has the following beneficial effects:

[0046] 1. The present application improves the comprehensiveness and representativeness of the collected samples by deploying distributed sampling nodes at multiple stages of cement production and combining multiple attributes such as particle size distribution, particle morphology and humidity to construct a weighted sample data cube. Compared with the existing static sample model, this method no longer relies on a limited sample library for static inference, but strengthens the dynamic association between samples and historical sieving behaviors through a representative weight mechanism. This method significantly improves the adaptability and generalization ability of the prediction model in cross-batch and multi-environment production scenarios, avoids prediction bias caused by single or distorted samples, and is suitable for high-frequency variable working condition requirements in actual working conditions.

[0047] 2. The present application proposes a screen state topology relationship modeling method based on a dynamic sieving log library, which can monitor and analyze key parameters such as sieve aperture flow and blockage degree in real time, and can identify the blockage trend and evolution path of the screen in a timely manner. This method introduces a simultaneous parameter solving and trend prediction mechanism, which not only improves the sensing accuracy of blockage events, but also actively triggers a compensation and calibration mechanism before blockage occurs. This mechanism enhances the sensing and response ability of the prediction model to changes in equipment state, effectively reducing the sieve residue misjudgment problem caused by abnormal equipment state.

[0048] 3、The present application constructs a coupling mapping model between environmental factors and particle motion behavior, comprehensively considers the influence of environmental variables such as temperature, humidity, air pressure and dust concentration on the screening process. When judging the abnormality of the screening result, the system can call the model to calibrate the environmental compensation of the screening prediction result, so as to correct the problems of particle aggregation, screen hole blockage and the like caused by environmental disturbance. Compared with the lack of dynamic modeling of environmental factors in the existing method, the method significantly improves the robustness and prediction reliability of the model under complex working conditions, and meets the accurate prediction demand under the high fluctuation working condition of the industrial field. BRIEF DESCRIPTION OF DRAWINGS

[0049] Figure 1 A flowchart of a cement negative pressure screen analyzer screen residue dynamic prediction method provided by an embodiment of the present application is shown in the figure.

[0050] Figure 2 A flowchart of generating a weighted sample data cube provided by an embodiment of the present application is shown in the figure.

[0051] Figure 3 A flowchart of constructing a screen mesh state topology relationship provided by an embodiment of the present application is shown in the figure.

[0052] Figure 4 A flowchart of generating a data correlation factor provided by an embodiment of the present application is shown in the figure. DETAILED DESCRIPTION

[0053] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0054] The prediction of the screen residue of the cement negative pressure screen analyzer is of great significance to ensure the stability of the quality of the cement product and the continuity of the production process. Through accurate prediction of the screen residue, not only can the change trend of the cement particle distribution be reflected in real time, and abnormal fluctuations in the production process can be found in time, but also the mill operation parameters and raw material ratio can be adjusted, and the rework and resource waste caused by unqualified screening can be effectively avoided. In addition, accurate prediction of the screen residue helps to improve the automation and intelligent level of screening detection, and provides key data support for realizing green, efficient and controllable cement production.

[0055] The present application proposes a cement negative pressure sieve analyzer residual amount dynamic prediction method to achieve high-precision residual amount prediction in complex production scenarios such as insufficient sample representation, frequent screen clogging and environmental factor disturbance. In order to illustrate that the method of the present application can play a role in high-precision residual amount prediction, the effectiveness of the present application will be illustrated from two embodiments below.

[0056] Embodiment one

[0057] In the embodiments of the present application, the high-precision residual amount prediction process of the method proposed by the present application in complex production scenarios such as insufficient sample representation, frequent screen clogging and environmental factor disturbance is described in detail. The embodiments of the present application are aimed at high-precision residual amount prediction in the cement production process of a certain factory A. The high-precision residual amount prediction process in the cement production process will be described in detail below according to the content; wherein, Figure 1 Figure 1 is a specific flowchart of the method proposed by the present application, including: constructing a production stage database, collecting a to-be-detected sample through a distributed node, and obtaining an attribute data table of the to-be-detected sample; establishing the association of the attribute data table and a historical sieve atlas by using association rule mining, and generating a weighted sample data cube; creating a dynamic sieve separation log library according to the weighted sample data cube, analyzing multi-dimensional parameters in the dynamic sieve separation log library through a data processing model, and constructing a screen state topology relationship; obtaining a sensor data stream through a multi-source sensor, calculating the similarity measurement of the sensor data stream and a particle distribution database by using a multi-dimensional index algorithm, and generating a data association factor; establishing a state discrimination rule library according to the screen state topology relationship and the data association factor, and performing prediction of the residual amount through a corresponding data processing method; performing weight vector aggregation on the first prediction value, the second prediction value and the third prediction value, and outputting a residual amount prediction result set. The following is explained in combination with the content in Figure 1 and Figure 2 .

[0058] A cement negative pressure sieve analyzer residual amount dynamic prediction method, comprising:

[0059] constructing a production stage database, collecting a to-be-detected sample through a distributed node, and obtaining an attribute data table of the to-be-detected sample; establishing the association of the attribute data table and a historical sieve atlas by using association rule mining, and generating a weighted sample data cube, as shown in Figure 2 .

[0060] The generation process of the weighted sample data cube includes:

[0061] dividing the cement production process into a plurality of production stages, and configuring N distributed sampling nodes for each production stage to collect to-be-detected samples;

[0062] ​Attribute data of each sample to be detected is acquired, including particle size distribution, particle morphology, particle spacing and humidity, an attribute data table is constructed, and is stored in a production stage database;

[0063] Potential correlations between the attribute data table and historical screening maps are analyzed using association rule mining to establish an association rule path.

[0064] Importance of each sample to be detected to a preset screening index is calculated according to the association rule path, and a representative weight is assigned to each sample to be detected;

[0065] The samples to be detected are combined by weighting according to the representative weight to generate a weighted sample data cube.

[0066] Specifically, the cement production process is divided into multiple stages, such as raw material batching, clinker calcination, mill grinding and finished product screening. N distributed sampling nodes (N = 10 in this embodiment) are configured for each stage, and samples to be detected are collected by automatic sampling equipment (such as a mechanical arm or a conveyor belt sampler). The sampling frequency is set to 2 times per hour to ensure that the samples cover the dynamic changes in production.

[0067] For each sample to be detected, a laser particle size analyzer, a microscope imaging system and a humidity sensor are used to measure particle size distribution, particle morphology, particle spacing and humidity, respectively. These attribute data are stored in a table form to generate an attribute data table, which is stored in a production stage database and managed by a relational database (such as MySQL).

[0068] The Apriori algorithm is used to analyze the potential correlations between the attribute data table and historical screening maps (containing the screening residue amount and screening conditions of historical samples). The historical screening maps are stored in the database and contain information such as screening residue amount, screen aperture, production parameters, etc. The mining process generates an association rule path.

[0069] According to the association rule path, the importance of each sample to be detected to a preset screening index (standard deviation of the screening residue amount in this embodiment) is calculated. The entropy method is used to calculate the weight: the information entropy of each attribute is calculated, and the lower the entropy value, the greater the impact of the attribute on the screening residue amount; the normalized entropy value is used to obtain a weight coefficient; a representative weight is assigned to each sample, and the weight value ranges from 0 to 1.

[0070] According to the representative weight, the attribute data of the samples to be detected are combined by weighting to generate a three-dimensional data cube, and the dimensions include sample ID, production stage and attribute type. The data cube is stored as a multi-dimensional array for subsequent analysis.

[0071] Table 1 is a comparison table of the prediction performance of the screening residue of the dynamic sample of the present application and the traditional static sample. Among them, the sample representativeness score is based on the feature distribution overlap of different sample sets (static sample and dynamic sample) and the full production data set, and the cross-batch adaptability score is based on the prediction accuracy difference of different sample models (static sample and dynamic sample) on different production batches not participating in training.

[0072] Table 1 is a comparison table of the prediction performance of the screening residue of the dynamic sample of the present application and the traditional static sample.

[0073] Predictive performance indicator Traditional static sample Dynamic sample of the present invention Cross-batch adaptability score 9.1 9.2 Sample representativeness score 4.6 5.1 Residual amount prediction error 3.4％ 9.7％

[0074] By generating a weighted sample data cube, the present method can effectively quantify the representativeness of the sample and reduce the prediction error caused by sample deviation. Compared with the traditional uniform sampling method, the present method improves the pertinence and information utilization rate of the sample data through association rule mining and weight distribution, so that the accuracy of the screening residue prediction is significantly improved in complex production scenarios.

[0075] Preferably, a dynamic screening log library is created according to the weighted sample data cube, and multi-dimensional parameters in the dynamic screening log library are analyzed by a data processing model to construct a screen state topology relationship, as shown in Figure 3

[0076] The construction process of the screen state topology relationship includes:

[0077] The weighted sample data cube is dynamically screened, and multi-dimensional parameters including screen hole flow, screen clogging degree and vibration frequency are monitored in real time to create a dynamic screening log library.

[0078] The multi-dimensional parameters are analyzed, the clogging rate is solved by simultaneous equations, and the clogging trend is identified and predicted.

[0079] The screen state topology relationship is constructed according to the clogging rate and the clogging trend.

[0080] Specifically, the weighted sample data cube is input into a negative pressure screen analyzer for dynamic screening, and multi-dimensional parameters including screen hole flow (mass of particles passing through the screen hole per unit time) and screen clogging degree (percentage of clogging area to total screen hole area) are monitored in real time. The monitoring equipment includes a flow sensor and an image analysis system. The monitoring data is stored in the form of time series to create a dynamic screening log library.

[0081] Based on the screen hole flow, the screen clogging degree and the vibration frequency, the following system of simultaneous equations is constructed:

[0082]

[0083] wherein, ​is the clogging rate; Q is the screen flow rate; B is the screen clogging degree; V is the vibration frequency; a and β are the adjustment coefficients. i and β i are the adjustment coefficients.

[0084] Different corresponding to the vibration frequency V are defined as:

[0085] Normal: and V > f min .

[0086] Clogging:

[0087] The evolution topology of "normal → clogging" is described in a graph database (such as Neo4j) in the form of nodes and directed edges, and the transition probability of each edge is stored.

[0088] By constructing a high-precision dynamic screening log library and combining a simultaneous equation model, the clogging degree of the screen and its evolution trend can be accurately reflected in real time, providing fine and traceable state information for subsequent screen residue prediction through a state discrimination rule library, thereby significantly improving the stability and reliability of screen residue prediction.

[0089] Preferably, the sensor data stream is obtained through a multi-source sensor, a multi-dimensional index algorithm is used to calculate the similarity measure of the sensor data stream and the particle distribution database, and a data correlation factor is generated, as shown in Figure 4 .

[0090] The generation process of the data correlation factor includes:

[0091] The sensor data stream obtained through the multi-source sensor is normalized, and the sensor data stream includes humidity, temperature, air pressure, and dust concentration.

[0092] A multi-dimensional feature vector is constructed based on the sensor data stream, and a multi-dimensional index algorithm is used to perform similarity matching on the particle distribution database to obtain a similarity measure value, and the K reference samples with the highest similarity measure value are extracted; the particle distribution database contains a plurality of paired samples of historical environmental characteristics and corresponding particle distribution curves.

[0093] In combination with the similarity measure value, the data correlation factor reflecting the relationship between the current environmental state and the historical particle distribution is generated.

[0094] Specifically, the sensor data stream including temperature, humidity, air pressure, and dust concentration is standardized by the min-max method or the Z-score method respectively, so as to ensure that each feature has the same dimension.

[0095] The normalized sensor data stream is spliced into a multi-dimensional feature vector. Each environmental feature-particle distribution curve sample in the historical particle distribution database is also pre-constructed as a four-dimensional feature vector set; a multi-dimensional R-tree structure is used to store the historical samples; during online query, the similarity of the multi-dimensional feature vector and the particle distribution database is quickly calculated based on Euclidean distance or cosine similarity, and a similarity ranking is performed to obtain a similarity value, and the K reference samples with the highest similarity value are extracted.

[0096] The K reference samples retrieved are weighted and averaged to obtain a data correlation factor; the weight is obtained from multiple tests.

[0097] Through efficient similarity comparison of multi-source sensor data and historical environmental particle distribution, the most representative reference sample can be dynamically extracted, and a quantifiable correlation factor can be formed, providing an accurate basis for environmental disturbance compensation and improving the adaptive ability of the prediction model to external changes.

[0098] Preferably, a data relationship model is called to output a first prediction value, and a state discrimination rule library is established according to the screen state topology relationship and the data correlation factor; the state discrimination rule library specifically includes:

[0099] When the clogging trend does not meet the preset clogging condition and the data correlation factor is lower than the preset correlation threshold, it is determined that the first rule is met;

[0100] When the clogging trend meets the preset clogging condition and the data correlation factor is lower than the preset correlation threshold, it is determined that the second rule is met;

[0101] When the clogging trend does not meet the preset clogging condition and the data correlation factor is higher than the preset correlation threshold, it is determined that the third rule is met.

[0102] Specifically, the data relationship model uses a machine learning algorithm (such as random forest or gradient boosting tree) to establish a mapping relationship between the attribute data table and the screen yield;

[0103] The attribute data table is standardized (mean = 0, variance = 1), and key features such as D50, circularity, humidity, etc. are extracted;

[0104] The random forest algorithm (number of trees = 100, maximum depth = 10) is used to minimize the mean square error;

[0105] The training process optimizes the model parameters through 5-fold cross-validation to ensure generalization ability.

[0106] The trained data relationship model is called to output the first prediction value by performing the following steps:

[0107] The attribute data table in the weighted sample data cube is input into the model, and multiple samples are batch processed to improve efficiency;

[0108] The model calculates a prediction value of the screen yield of each sample according to the input key features.

[0109] The prediction values are weighted and averaged according to the representative weights of the samples to generate a first prediction value.

[0110] The specific discrimination conditions of the state discrimination rule base include:

[0111] The first rule (slight disturbance):

[0112] Condition: clogging rate and data correlation factor <0.3;

[0113] Decision: directly output the first prediction value.

[0114] The second rule (clogging dominant):

[0115] Condition: clogging rate and data correlation factor <0.3;

[0116] Decision: perform first compensation calibration, output second prediction value after adjusting clogging deviation.

[0117] The third rule (environmental disturbance):

[0118] Condition: clogging rate and data correlation factor >0.3;

[0119] Decision: perform second compensation calibration, output third prediction value through environmental impact model.

[0120] Based on the branch rule discrimination of different dominant factors (mechanical clogging vs. environmental disturbance), different production environments can be accurately selected for prediction and calibration using targeted models, overcoming the problem of insufficient robustness of a single model, improving the self-adaptation of the environment, and thus improving the overall prediction accuracy.

[0121] Preferably, when the first rule is met, the first prediction value is directly output; when the second rule is met, the first compensation calibration is performed based on the screen state topological relationship to generate the second prediction value.

[0122] The specific process of the first compensation calibration includes:

[0123] The deviation influence of the multi-dimensional parameters on the screen yield prediction is analyzed in combination with the screen state topological relationship.

[0124] According to the deviation influence, a compensation amount for correcting the second prediction value is calculated.

[0125] The calculated compensation amount is combined with the first predicted value to generate a second predicted value after first compensation calibration.

[0126] Specifically, when the second rule is triggered, multi-dimensional parameter deviation within the last 10 minutes is extracted;

[0127] A compensation function is defined based on the multi-dimensional parameter deviation:

[0128] C=k1·(ΔQ) 2 +k2·(ΔB) 2 +k3·(ΔV) 2 +k4·ΔQ·ΔB+k5·ΔQ·ΔV+k6·ΔB·ΔV;

[0129] Wherein, C is the compensation amount; ΔQ is the screen opening flow deviation; ΔB is the screen clogging degree deviation; ΔV is the vibration frequency deviation; k1, k2, k3, k4, k5 and k6 are weights;

[0130] The compensation amount is added to the first predicted value to generate a second predicted value.

[0131] Table 2 is a comparison table of the ability of introducing clogging compensation to screen residue prediction.

[0132] Table 2 Comparison table of the ability of clogging compensation to screen residue prediction

[0133] Predictive performance indicator Without introducing jam compensation With introducing jam compensation Jam early identification rate (%) Unable to identify early 90.6 Compensation trigger timeliness (ms) Unable to trigger compensation 38 Residual amount prediction error (%) 25.3 4.6

[0134] The first compensation calibration corrects the model deviation by quantitative analysis and real-time compensation of clogging deviation in the screening process, reduces the system error caused by clogging, and realizes more reliable near-time prediction.

[0135] Preferably, when the third rule is met, a second compensation calibration is implemented through an environmental impact model to generate a third predicted value;

[0136] The specific process of the second compensation calibration includes:

[0137] Through the environmental impact model, a coupling mapping relationship between environmental factors and particle motion characteristics is constructed, including temperature-diffusion rate function, humidity-aggregation coefficient function, air pressure-particle migration probability function and dust concentration-screen opening permeability function;

[0138] The coupling mapping relationship is applied to the current sensor data stream to dynamically calculate an environmental interference influence coefficient matrix;

[0139] The particle distribution model is corrected according to the influence coefficient matrix to inverse the screen residue deviation trend in the screening process;

[0140] The first prediction value is adjusted in combination with the inversion result to generate a third prediction value which is compensated and calibrated.

[0141] Specifically, the temperature-diffusion rate, humidity-aggregation coefficient, air pressure-migration probability, and dust concentration-permeability are respectively experimentally fitted to obtain the coupling mapping relationship.

[0142] The coupling mapping relationship is applied to the current sensor data stream to calculate an influence coefficient matrix, and the matrix elements represent the weights of the environmental factors on the particle distribution. The matrix has a dimension of 4xN (N is the number of particle distribution characteristics), and the matrix decomposition algorithm is used to optimize the calculation efficiency.

[0143] The particle distribution model is corrected according to the influence coefficient matrix, and the Monte Carlo simulation is used to invert the screen residue deviation trend in the screening process to obtain the deviation amount.

[0144] The first prediction value is adjusted to generate a third prediction value.

[0145] Table 3 is a comparison table of screen residue prediction performance under environmental disturbance.

[0146] The environmental fluctuation index is obtained by calculating the standard deviation and fluctuation coefficient of each environmental parameter in the current sensor data stream, and performing z-score standardization and weighted processing on the standard deviation and fluctuation coefficient.

[0147] According to the environmental fluctuation index, the environmental fluctuation intensity is classified as follows: the environmental fluctuation index is between 0 and 1, which is middle-high level; the environmental fluctuation index is between 1 and 2, which is high level; and the environmental fluctuation index is greater than 2, which is extremely high level.

[0148] Table 3 is a comparison table of screen residue prediction performance under environmental disturbance.

[0149] Environmental fluctuation intensity level Existing method prediction error (%) Invention method prediction error (%) High 11.4 4.8 High 10.2 4.2 Extremely high 14.7 3.9 High 12.1 4.7 Extremely high 15.3 4.1

[0150] The second compensation calibration introduces environmental coupling mapping, which effectively quantifies the comprehensive influence of multiple environmental factors on particle motion and screening efficiency by coupling environmental factors and particle motion characteristics, so that the prediction model has good environmental adaptability and generalization ability.

[0151] Preferably, the first prediction value, the second prediction value and the third prediction value are aggregated by a weight vector to output a screen residue prediction result set.

[0152] The output process of the screen residue prediction result set includes:

[0153] Integrating the first prediction value, the second prediction value and the third prediction value generated according to different rules;

[0154] Weights are assigned based on reliability and importance of each group of prediction values, and the three groups of prediction values are weighted and fused to obtain a comprehensive screen residue prediction value.

[0155] The comprehensive screen residue prediction value is output in a structured format as a final screen residue prediction result set.

[0156] Specifically, weights are assigned based on reliability and importance of each group of prediction values, the reliability is obtained by calculating the inverse of historical prediction error, and the importance is determined according to a rule priority;

[0157] The first prediction value, the second prediction value and the third prediction value are weighted and summed according to the assigned weights to obtain a comprehensive screen residue prediction value.

[0158] Through multi-model weighted fusion, the advantages of each model are complementary, the jamming calibration and environmental compensation are comprehensively considered, the fusion of multi-scene prediction results is realized, and finally a screen residue prediction result with higher accuracy and stability is output.

[0159] Embodiment two

[0160] In embodiment one, the method successfully realizes high-precision screen residue prediction in complex production scenarios such as insufficient sample representation, frequent screen jamming and environmental factor disturbance. In order to further verify the effectiveness of the present application, high-precision screen residue prediction is also carried out in the cement production process of a certain factory B in the embodiment of the present application.

[0161] A cement negative pressure screen analyzer screen residue dynamic prediction method, comprising:

[0162] A production stage database is constructed, a to-be-detected sample is collected by a distributed node, and an attribute data table of the to-be-detected sample is obtained; the association between the attribute data table and a historical screen atlas is established by using association rule mining, and a weighted sample data cube is generated;

[0163] The generation process of the weighted sample data cube comprises:

[0164] The cement production process is divided into multiple production stages, and N distributed sampling nodes are configured for each production stage to collect to-be-detected samples;

[0165] The attribute data of each to-be-detected sample is obtained, including particle size distribution, particle morphology, particle spacing and humidity, an attribute data table is constructed, and stored in the production stage database;

[0166] The potential association between the attribute data table and the historical screen atlas is analyzed by using association rule mining, and an association rule path is established;

[0167] Calculate the importance of each to-be-detected sample to the preset screening index according to the association rule path, and assign a representative weight to each to-be-detected sample;

[0168] According to the representative weight, the to-be-detected samples are combined by weighting to generate the weighted sample data cube.

[0169] Preferably, a dynamic screening log library is created according to the weighted sample data cube, and multi-dimensional parameters in the dynamic screening log library are analyzed by a data processing model to construct a screen state topology relationship;

[0170] The construction process of the screen state topology relationship includes:

[0171] The weighted sample data cube is dynamically screened to monitor multi-dimensional parameters in real time, including screen hole flow, screen clogging degree and vibration frequency, and a dynamic screening log library is created;

[0172] The multi-dimensional parameters are analyzed, the clogging rate is solved by simultaneous equations, and the clogging trend is identified and predicted;

[0173] The screen state topology relationship is constructed according to the clogging rate and the clogging trend.

[0174] Preferably, a multi-source sensor is used to obtain a sensor data stream, a multi-dimensional index algorithm is used to calculate the similarity measure of the sensor data stream and the particle distribution database, and a data correlation factor is generated;

[0175] The generation process of the data correlation factor includes:

[0176] The sensor data stream obtained by the multi-source sensor is normalized, and the sensor data stream includes humidity, temperature, air pressure and dust concentration;

[0177] Based on the sensor data stream, a multi-dimensional feature vector is constructed, a multi-dimensional index algorithm is used to perform similarity matching on the particle distribution database to obtain a similarity measure value, and the K reference samples with the highest similarity measure value are extracted; the particle distribution database contains a plurality of paired samples of historical environmental characteristics and corresponding particle distribution curves;

[0178] In combination with the similarity measure value, the data correlation factor reflecting the relationship between the current environmental state and the historical particle distribution is generated.

[0179] Preferably, a data relationship model is called to output a first prediction value, and a state discrimination rule library is established according to the screen state topology relationship and the data correlation factor;

[0180] The state discrimination rule library specifically includes:

[0181] When the clogging trend does not satisfy the preset clogging condition and the data correlation factor is lower than the preset correlation threshold, it is determined that the first rule is met;

[0182] When the clogging trend satisfies the preset clogging condition and the data correlation factor is lower than the preset correlation threshold, it is determined that the second rule is met;

[0183] When the clogging trend does not satisfy the preset clogging condition and the data correlation factor is higher than the preset correlation threshold, it is determined that the third rule is met.

[0184] Preferably, when the first rule is met, a first predicted value is output by calling a data relationship model; when the second rule is met, a second predicted value is generated by performing a first compensation calibration based on a screen state topology relationship;

[0185] The specific process of the first compensation calibration includes:

[0186] In combination with the screen state topology relationship, the deviation influence of the multi-dimensional parameters on the screen residue prediction is analyzed;

[0187] According to the deviation influence, a compensation amount for correcting the second predicted value is calculated;

[0188] The calculated compensation amount is combined with the first predicted value to generate the second predicted value after the first compensation calibration.

[0189] Preferably, when the third rule is met, a third predicted value is generated by implementing a second compensation calibration through an environmental impact model;

[0190] The specific process of the second compensation calibration includes:

[0191] Through the environmental impact model, a coupling mapping relationship between environmental factors and particle motion characteristics is constructed, including a temperature-diffusion rate function, a humidity-aggregation coefficient function, an air pressure-particle migration probability function, and a dust concentration-screen hole permeability function;

[0192] The coupling mapping relationship is applied to the current sensor data stream to dynamically calculate an environmental disturbance influence coefficient matrix;

[0193] According to the influence coefficient matrix, a particle distribution model is corrected to inverse the screen residue deviation trend in the screening process;

[0194] In combination with the inversion result, the first predicted value is adjusted to generate the third predicted value after the second compensation calibration.

[0195] Preferably, the first predicted value, the second predicted value, and the third predicted value are aggregated by a weight vector to output a screen residue prediction result set.

[0196] The output process of the screen residue prediction result set includes:

[0197] integrating the first, second and third predicted values generated according to different rules;

[0198] assigning weights based on the reliability and importance of each set of predicted values, and performing weighted fusion on the three sets of predicted values to obtain a comprehensive screen undersize prediction value;

[0199] outputting the comprehensive screen undersize prediction value in a structured format as the final screen undersize prediction result set.

[0200] Table 4 is a comparison table of screen undersize prediction performance of the overall scheme of the present application.

[0201] wherein the sample representativeness score is obtained by calculating the proportion of the effective coverage dimension of the sample space in the theoretical feature space dimension;

[0202] The multi-working-condition adaptability score is obtained by calculating the cross-working-condition error mean;

[0203] The screen clogging perception accuracy score is evaluated by the clogging recognition accuracy;

[0204] The environmental disturbance compensation effectiveness score is obtained by comparing the prediction error difference before and after compensation;

[0205] The overall prediction accuracy score is calculated using the standard MAE formula.

[0206] Table 4 is a comparison table of screen undersize prediction performance evaluation of the overall scheme of the present application

[0207] Evaluation indicator Existing method Invention scheme Sample representativeness score 5.2 9.1 Multi-working-condition adaptability score 4.9 8.8 Screen jam perception accuracy score 5.7 9.2 Environmental disturbance compensation effectiveness score 4.8 8.9 Overall prediction accuracy score 5.3 9.3

[0208] Although embodiments of the present application have been shown and described, it is to be understood that various modifications, substitutions, replacements and changes can be made to these embodiments without departing from the principles and spirit of the present application, and the scope of the present application is defined by the appended claims and their equivalents.

Claims

1. A method for dynamically predicting the residue on a cement negative pressure sieve analyzer, characterized in that, include: Build a production phase database, collect samples to be tested through distributed nodes, and obtain the attribute data table of the samples to be tested; By using association rule mining, the correlation between the attribute data table and the historical screening map is established, and a weighted sample data cube is generated. A dynamic screening log library is created based on a weighted sample data cube. Multi-dimensional parameters in the dynamic screening log library are analyzed through a data processing model to construct the topological relationship of the screening state. By acquiring sensor data streams from multiple sources, a multi-dimensional indexing algorithm is used to calculate the similarity measure between the sensor data streams and the particle distribution database, and a data association factor is generated. The first predicted value is output through the data relationship model, and a state discrimination rule base is established based on the topological relationship of the screen state and the data association factor: when the first rule is met, the first predicted value is directly output. When the second rule is met, a first compensation calibration is performed based on the topological relationship of the screen state to generate a second predicted value; When the third rule is met, a second compensation calibration is performed using the environmental impact model to generate a third predicted value; The weighted vectors of the first, second, and third predicted values ​​are aggregated to output a set of prediction results for the remaining amount of sieve.

2. The method for dynamically predicting the residue of a cement negative pressure sieve analyzer according to claim 1, characterized in that, The process of generating the weighted sample data cube includes: The cement production process is divided into multiple production stages, and N distributed sampling nodes are configured for each production stage to collect samples to be tested. Acquire the attribute data of each sample to be tested, including particle size distribution, particle morphology, particle spacing and humidity, construct an attribute data table and store it in the production stage database; Utilize association rules to mine and analyze the potential relationships between attribute data tables and historical screening maps, and establish association rule paths; The importance of each sample to be detected to the preset screening index is calculated based on the association rule path, and a representative weight is assigned to each sample to be detected. The samples to be tested are weighted and combined according to the representative weights to generate the weighted sample data cube.

3. The method for dynamically predicting the residue of a cement negative pressure sieve analyzer according to claim 1, characterized in that, The process of constructing the topological relationship of the screen state includes: The weighted sample data cube is dynamically sieved, and multi-dimensional parameters, including sieve flow rate, sieve clogging degree and vibration frequency, are monitored in real time to create a dynamic sieve log library. The multi-dimensional parameters are analyzed, the clogging rate is solved by solving simultaneous equations, and the clogging trend is identified and predicted. The screen state topology is constructed based on the clogging rate and the clogging trend.

4. The method for dynamically predicting the residue of a cement negative pressure sieve analyzer according to claim 1, characterized in that, The process of generating the data association factor includes: The sensor data stream acquired by the multi-source sensors is normalized, and the sensor data stream includes humidity, temperature, air pressure and dust concentration. A multidimensional feature vector is constructed based on the sensor data stream, and a multidimensional indexing algorithm is used to perform similarity matching on the particle distribution database to obtain a similarity metric value. The K reference samples with the highest similarity metric values ​​are extracted. The particle distribution database contains paired samples of multiple historical environmental features and corresponding particle distribution curves. By combining the similarity metric, a data association factor reflecting the relationship between the current environmental state and historical particle distribution is generated.

5. The method for dynamically predicting the residue of a cement negative pressure sieve analyzer according to claim 1, characterized in that, The state discrimination rule base specifically includes: When the congestion trend does not meet the preset congestion conditions and the data correlation factor is lower than the preset correlation threshold, it is determined to meet the first rule; When the congestion trend meets the preset congestion conditions and the data correlation factor is lower than the preset correlation threshold, it is determined to meet the second rule; When the congestion trend does not meet the preset congestion conditions and the data correlation factor is higher than the preset correlation threshold, it is determined to meet the third rule.

6. The method for dynamically predicting the residue of a cement negative pressure sieve analyzer according to claim 1, characterized in that, The specific process of the first compensation calibration includes: Based on the topological relationship of the screen state, the influence of the multi-dimensional parameters on the deviation of the screen residue prediction is analyzed; Based on the aforementioned deviation effect, calculate the compensation amount used to correct the second predicted value; The calculated compensation amount is combined with the first predicted value to generate a second predicted value that has been calibrated by the first compensation.

7. The method for dynamically predicting the residue of a cement negative pressure sieve analyzer according to claim 1, characterized in that, The specific process of the second compensation calibration includes: Through environmental impact models, a coupled mapping relationship between environmental factors and particle motion characteristics is constructed, including temperature-diffusion rate function, humidity-aggregation coefficient function, air pressure-particle migration probability function, and dust concentration-sieve permeability function. The coupling mapping relationship is applied to the current sensing data stream to dynamically calculate the environmental interference influence coefficient matrix. Based on the influence coefficient matrix, the particle distribution model is corrected, and the trend of sieve residue shift during the screening process is inverted. The first predicted value is offset and adjusted based on the inversion results to generate a third predicted value after second compensation calibration.

8. The method for dynamically predicting the residue of a cement negative pressure sieve analyzer according to claim 1, characterized in that, The output process of the screening residue prediction result set includes: integrating the first prediction value, the second prediction value, and the third prediction value generated according to different rules; assigning weights based on the reliability and importance of each group of prediction values, and performing weighted fusion of the three groups of prediction values ​​to obtain a comprehensive screening residue prediction value; and outputting the comprehensive screening residue prediction value in a structured format as the final screening residue prediction result set.

Images