Burning-through point prediction method based on big data
By combining principal component analysis and support vector machine algorithms with a big data platform, accurate prediction of the sintering endpoint was achieved, solving the problems of insufficient subjectivity and real-time performance in traditional methods, improving production efficiency and product quality, and promoting intelligent management.
Patent Information
- Application Number
- CN202511909766.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-17
- Publication Date
- 2026-05-26
AI Technical Summary
Traditional methods for determining the sintering endpoint rely on manual monitoring or simple physical parameters, which are subjective and cannot reflect sudden changes in the production process in a timely manner, resulting in low production efficiency, energy waste, and unstable product quality.
By integrating principal component analysis (PCA) and support vector machine (SVM) algorithms using a big data platform and combining them with process algorithms, the system can monitor and accurately predict key locations in the sintering process in real time and provide expert advice.
It has improved the quality and stability of sintered ore, reduced energy consumption, increased production efficiency and product quality, and promoted intelligent management of industrial production.
Smart Images
Figure CN122087554A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to a sintering endpoint prediction method based on big data, which is particularly applicable to industrial production such as metallurgy and steel. By using big data analysis technology to monitor and analyze the sintering process in real time, the sintering endpoint can be accurately predicted. Background Technology
[0002] Sintering is a crucial process in steel production, and the quality of the sinter directly impacts the efficiency of subsequent smelting processes. The endpoint of sintering is related to many production parameters, such as: bed thickness, sintering machine speed, raw material ratio, bellows negative pressure, bellows temperature, and the airflow and opening of the main exhaust fan. An early endpoint indicates under-sintering, meaning the sintering machine's capacity is not fully utilized, resulting in reduced sinter production. Conversely, a delayed endpoint indicates over-sintering, leading to excessive energy consumption. This is because the furnace temperature remains high after sintering, and continued heating and sintering wastes significant energy, reducing production efficiency and increasing costs.
[0003] Precise control of the sintering endpoint is crucial for improving sintering quality, reducing energy consumption, and minimizing environmental pollution. However, traditional methods for determining the sintering endpoint mainly rely on manual monitoring or simple physical parameters such as temperature and atmosphere, using experience to judge the end time of the sintering process. Such methods are not only subjective but also unable to reflect sudden changes that occur during production. Therefore, a more accurate and intelligent prediction method is needed. Summary of the Invention
[0004] To address the aforementioned issues, this invention provides a big data-based method for predicting the sintering endpoint. By integrating a big data platform and using principal component analysis (PCA), support vector machine (SVM), and process algorithms, it achieves real-time monitoring and accurate prediction of key positions such as TRP, BRP, and BTP during the sintering process, and provides reasonable expert advice, significantly improving the quality and production stability of sintered ore.
[0005] To achieve the above objectives, the present invention provides a sintering endpoint prediction method based on big data, comprising the following steps: Collect historical data during the sintering process; The collected historical data is cleaned and preprocessed; Effective features are extracted from preprocessed historical data, and the most predictive input variables are determined using feature selection methods. A sintering endpoint prediction model is trained based on the extracted features using the support vector machine (SVM) algorithm. Based on the trained prediction model, the sintering endpoint is predicted.
[0006] Furthermore, the historical data during the sintering process includes: ingredient ratio, moisture content of the first and second mixtures, material layer thickness, ignition furnace temperature, air pressure and flow rate, gas pressure and flow rate, sintering machine speed, inlet pressure and temperature of the main flue, negative pressure and temperature of the wind box, main exhaust volume and damper opening.
[0007] Furthermore, after the steps of cleaning and preprocessing the collected data, the method further includes: performing differential processing on the time series data to eliminate trends and seasonality in the data, making the data more stable and beneficial for subsequent modeling and analysis.
[0008] Furthermore, the feature selection methods include variable importance ranking, correlation analysis, or backward selection.
[0009] Furthermore, the sintering endpoint prediction includes: the prediction of the temperature rise point (TRP), the sintering inflection point (BRP), and the sintering endpoint (BTP) during the sintering process.
[0010] Furthermore, it also includes a step of providing real-time feedback based on the prediction results.
[0011] Furthermore, the preprocessing steps include: (1) Data cleaning: Remove invalid or abnormal data points, such as missing values, duplicate records, etc.
[0012] (2) Noise removal: The noise caused by sensor error, environmental factors or system fluctuations is removed by filtering algorithm to improve data quality.
[0013] (3) Normalization: Standardize the data of different dimensions and scales so that the feature data can be compared on the same scale, avoiding unnecessary impact on subsequent analysis due to differences in dimensions.
[0014] Preprocessed data will ensure the consistency and high quality of input data during model training, thereby improving the accuracy of the prediction model.
[0015] Furthermore, the step of extracting effective features includes: (1) Data standardization: Standardize each feature to make it have the same scale and avoid the influence of differences in units; (2) Covariance matrix calculation: Calculate the covariance matrix of standardized data to reflect the correlation between features; (3) Eigenvalue decomposition: Perform eigenvalue decomposition on the covariance matrix to obtain eigenvectors and corresponding eigenvalues. The eigenvalues represent the variability of the data in that direction. (4) Principal component selection: Based on the magnitude of the eigenvalues, select the top k eigenvectors with the greatest variability as principal components; (5) Dimensionality reduction: Map the original data onto the selected principal components to reduce the data dimensionality; Principal components extracted by PCA are used as input features to remove redundancy and noise, reduce computational complexity, and improve model prediction accuracy and stability. Furthermore, the PCA algorithm improves data processing efficiency through dimensionality reduction, enhancing the model's robustness.
[0016] The present invention has the following advantages: 1. Accurate prediction: This invention uses big data technology and machine learning algorithms to comprehensively analyze sintering process parameters, thereby achieving accurate prediction of the sintering endpoint, which greatly improves the accuracy and reliability of the prediction.
[0017] 2. Improve production efficiency: By predicting the sintering endpoint in advance, operators can better control the production rhythm and avoid ending the sintering process too early or too late, thereby improving production efficiency and reducing energy consumption.
[0018] 3. Optimize product quality: Accurate endpoint prediction helps control key parameters in the sintering process, ensuring stable sinter quality and improving the efficiency and output of subsequent smelting processes.
[0019] 4. Intelligent Management: This invention utilizes big data analysis and machine learning technology to provide intelligent decision support for the sintering process, promoting the development of industrial production towards intelligence and automation. Attached Figure Description
[0020] Figure 1 This is a flowchart illustrating the principle of this invention.
[0021] Figure 2 This is the bellows temperature diagram of the present invention.
[0022] Figure 3 This is the past, actual, and predicted TRP, BRP, and BTP diagram of the present invention.
[0023] Figure 4 This is the sintering endpoint BTP soft measurement and prediction diagram of the present invention. Detailed Implementation
[0024] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings.
[0025] In the description of this invention, it should be understood that the terms "center", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.
[0026] The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, unless otherwise stated, "a plurality of" means two or more.
[0027] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "joining" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal communication between two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.
[0028] Figure 1 The figure shows the sintering endpoint prediction method based on big data of the present invention, which was tested in a 240m2 sintering machine in a steel plant.
[0029] First, a big data sintering platform is constructed. This platform collects data extensively and in-depth based on industrial instruments and meters, building a sintering big data platform centered on equipment data from the production site. It accesses different devices, systems, and products through various communication methods to collect massive amounts of data; simultaneously, it utilizes protocol conversion technology to normalize multi-source heterogeneous data; and the platform establishes unified data and service interface specifications to facilitate rapid program development and application.
[0030] The system integrates and consolidates parameters from the entire sintering process, providing users with comprehensive data services and ensuring the authenticity, systematic nature, and uniqueness of the big data system. The specific development process is as follows: (1) Network debugging: Ensure normal communication between all nodes and ports of the big data server, and ensure network interconnection with the data acquisition server, big data platform application server, and client. (2) The big data server cluster uses the Linux system, and the system ensures passwordless login, clock synchronization, Java, database and other environment setup.
[0031] (3) Deploy components such as HDFS, Hive, HBase, YARN, ZooKeeper, Flume, Flink, Kafka, and Impala on the big data server cluster.
[0032] (4) The big data platform is developed using the Spring Boot + Vue + MySQL + ActiveMQ + Nginx environment, and the environment is deployed on the big data platform application server.
[0033] (5) Data access to the big data platform adopts data access methods such as OPC UA, HTTP, ActiveMQ, Kafka, manual import, and script import. Data is classified and stored on HDFS. Hive is used to perform large-scale analysis and processing of the stored data. HBase is used to store massive amounts of time-series data and provide data support for the big data platform application server.
[0034] 1. Collect historical data This invention collects production process data from the sintering site using KepServer. The data originates from equipment within the primary automation system, including but not limited to PLCs and DCS control systems. The collected historical data covers multiple key process parameters, including raw material ratios, moisture content and temperature during the first and second mixing stages, sintering machine speed, roller speed, nine-roller speed, main flue negative pressure and temperature, wind box negative pressure and temperature, and the airflow and damper opening of the main exhaust fan. This data reflects each stage of the sintering process, from raw material preparation to the sintering machine's operating status, providing crucial input information for the predictive model. The collected historical data is connected to a big data platform via an appropriate interface for subsequent analysis and modeling.
[0035] 2. Data preprocessing Comprehensive preprocessing of the collected historical data is performed on a big data platform to ensure data accuracy and consistency. Preprocessing steps include: (1) Data cleaning: Remove invalid or abnormal data points, such as missing values, duplicate records, etc.
[0036] (2) Noise removal: The noise caused by sensor error, environmental factors or system fluctuations is removed by filtering algorithm to improve data quality.
[0037] (3) Normalization: Standardize the data of different dimensions and scales so that the feature data can be compared on the same scale, avoiding unnecessary impact on subsequent analysis due to differences in dimensions.
[0038] Preprocessed data will ensure the consistency and high quality of input data during model training, thereby improving the accuracy of the prediction model.
[0039] 3. Serial difference To eliminate trends and seasonality in historical data, sequential differencing is employed. By differencing each feature, long-term trends and periodic fluctuations are eliminated, making the data more stable and suitable for modeling analysis. Sequential differencing helps reduce interference from external factors, enhancing the predictability of the data and providing a more stable foundation for subsequent modeling analysis. Data processed through differencing exhibits better stability, facilitating subsequent model training and prediction.
[0040] 4. Feature extraction and selection; This invention utilizes Principal Component Analysis (PCA) algorithm for feature extraction and selection, extracting representative features from preprocessed data to improve the accuracy and efficiency of the model. Specific steps include: (1) Data standardization: Standardize each feature to make it have the same scale and avoid the influence of differences in dimensions.
[0041] (2) Covariance matrix calculation: Calculate the covariance matrix of standardized data to reflect the correlation between features.
[0042] (3) Eigenvalue decomposition: Eigenvalue decomposition is performed on the covariance matrix to obtain eigenvectors and corresponding eigenvalues. The eigenvalues represent the variability of the data in that direction.
[0043] (4) Principal component selection: Based on the magnitude of the eigenvalues, select the top k eigenvectors with the greatest variability as principal components.
[0044] (5) Dimensionality reduction: Map the original data onto the selected principal components to reduce the data dimensionality.
[0045] Principal components extracted by PCA are used as input features to remove redundancy and noise, reduce computational complexity, and improve model prediction accuracy and stability. Furthermore, the PCA algorithm improves data processing efficiency through dimensionality reduction, enhancing the model's robustness.
[0046] 5. Use the Support Vector Machine (SVM) algorithm to build and train the model; This invention employs the Support Vector Machine (SVM) algorithm to construct a sintering endpoint prediction model. By finding an optimal hyperplane, the distance between data points and the hyperplane is maximized, thereby enabling classification or regression analysis.
[0047] First, the preprocessed 86,000 historical samples were randomly divided into training and testing samples. 62,000 samples were used as training samples to train the model parameters and adjust the hyperparameters of the support vector machine to ensure optimal model fit on the training data. The remaining 24,000 samples were used as testing samples to verify the model's predictive performance and generalization ability.
[0048] Through iterative optimization algorithms, the hyperparameters of the model (such as C-value and kernel function) are repeatedly adjusted. Cross-validation and error analysis are then used to arrive at an optimal prediction model. Furthermore, a 5-minute prediction period is selected, which has been empirically verified to accurately reflect the dynamic changes during the sintering process, providing timely and effective decision-making support for production scheduling and quality control.
[0049] 6. TRP, BRP, and BTP prediction; Reference Figure 2 During a specific period, we obtained data on the stable production process of the sintering machine at the steel plant and conducted a detailed analysis of the exhaust gas temperature data from air boxes 1 to 24. The steel plant has 24 air boxes, and the normal endpoint of the sintering process is typically located between air boxes 22 and 23. The total length of the trolley is 90 meters, and we selected 82 meters as the benchmark for the endpoint. If the actual endpoint is less than 82 meters, it is defined as "early," i.e., "under-sintering"; if it is greater than 82 meters, it is considered "late," i.e., "over-sintering." This standard helps us accurately assess the progress of the sintering process and take appropriate adjustment measures.
[0050] In analyzing temperature data, we often find that the highest temperature point corresponds to the end point of sintering. Through careful observation and analysis of this temperature data, we can gain key insights into the sintering process. This temperature distribution analysis provides an important basis for determining the sintering endpoint, helping us accurately assess sintering quality and ensure the stability of the production process and the high quality of the product.
[0051] Reference Figure 3 Based on the pre-trained model, we can accurately predict the location and state of various key points in the sintering process, including the temperature rise point (TRP), the sintering inflection point (BRP), and the sintering endpoint (BTP). The past values and states represent the previous prediction data, i.e., the data from 5 minutes ago. These predictions not only provide a basis for real-time monitoring of the production process but also help us make more accurate decisions during sintering, thereby optimizing production efficiency and quality control.
[0052] 7. Results feedback and suggestions; Reference Figure 4We randomly selected a day for analysis. By observing the data within this period, we found that the soft measurement value of the sintering endpoint (BTP) consistently maintained a high degree of consistency with the predicted value. This result demonstrates that the adopted prediction model possesses high accuracy and stability in actual production, effectively reflecting the endpoint changes during sintering, further proving the model's reliability and effectiveness. This consistency not only enhances the predictive ability of the sintering process but also provides strong support for real-time adjustment and optimization of production parameters.
[0053] Regarding the predicted sintering endpoint (BTP) location and state given in step seven, we can provide a series of expert suggestions based on the actual situation. For example, if the prediction shows a delayed sintering endpoint, it may indicate a risk of over-firing during the sintering process. In this case, we recommend the following process adjustment measures: (1) Reduce the speed of the sintering machine: By appropriately reducing the operating speed of the sintering machine, the residence time of the material in each air box is optimized, thereby avoiding over-burning.
[0054] (2) Reduce the thickness of the material layer: Appropriately reducing the thickness of the material layer helps to improve the heat transfer efficiency, reduce the accumulation of heat in the material layer, and prevent overburning. In addition, adjusting the thickness of the material layer can also effectively control the temperature distribution and ensure the uniformity of the sintering process.
[0055] These adjustments help on-site operators respond promptly to changes during the sintering process, ensuring stable sintering quality and efficient production. These recommendations provide workers with valuable guidance, enabling them to make quick and accurate decisions in complex production environments.
[0056] The present invention has been described in detail above with reference to the accompanying drawings. However, the present invention is not limited to the embodiments described above. Within the scope of knowledge possessed by those skilled in the art, various changes can be made without departing from the spirit of the present invention. Many other changes and modifications made without departing from the concept and scope of the present invention should be considered within the scope of protection of the present invention.
[0057] In the description of this specification, specific features, structures, materials, or characteristics may be combined in any suitable manner in one or more embodiments or examples.
[0058] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.
Claims
1. A sintering endpoint prediction method based on big data, characterized in that, The method includes the following steps: Collect historical data during the sintering process; The collected historical data is cleaned and preprocessed; Effective features are extracted from preprocessed historical data, and the most predictive input variables are determined using feature selection methods. A sintering endpoint prediction model is trained based on the extracted features using the support vector machine (SVM) algorithm. Based on the trained prediction model, the sintering endpoint is predicted.
2. The sintering endpoint prediction method based on big data as described in claim 1, characterized in that, The historical data during the sintering process includes: ingredient ratio, moisture content of the first and second mixtures, material layer thickness, ignition furnace temperature, air pressure and flow rate, gas pressure and flow rate, sintering machine speed, inlet pressure and temperature of the main flue, negative pressure and temperature of the wind box, main exhaust volume and damper opening.
3. The sintering endpoint prediction method based on big data as described in claim 1, characterized in that, After the steps of cleaning and preprocessing the collected data, the method further includes: performing differential processing on the time series data to eliminate trends and seasonality in the data, making the data more stable and beneficial for subsequent modeling and analysis.
4. The sintering endpoint prediction method based on big data as described in claim 1, characterized in that, The feature selection methods mentioned include variable importance ranking, correlation analysis, or backward selection.
5. The sintering endpoint prediction method based on big data as described in claim 1, characterized in that, The sintering endpoint prediction includes the prediction of the temperature rise point (TRP), the sintering inflection point (BRP), and the sintering endpoint (BTP) during the sintering process.
6. The sintering endpoint prediction method based on big data as described in claim 1, characterized in that, It also includes a step of providing real-time feedback based on the prediction results.
7. The sintering endpoint prediction method based on big data as described in claim 1, characterized in that, The preprocessing steps include: (1) Data cleaning: removing invalid or abnormal data points; (2) Noise removal: Noise caused by sensor error, environmental factors or system fluctuations is removed by filtering algorithm to improve data quality; (3) Normalization: Standardize the data of different dimensions and scales so that the feature data can be compared on the same scale, avoiding unnecessary impact on subsequent analysis due to differences in dimensions.
8. The sintering endpoint prediction method based on big data as described in claim 1, characterized in that, The steps for extracting effective features include: (1) Data standardization: Standardize each feature to make it have the same scale and avoid the influence of differences in units; (2) Covariance matrix calculation: Calculate the covariance matrix of standardized data to reflect the correlation between features; (3) Eigenvalue decomposition: Perform eigenvalue decomposition on the covariance matrix to obtain eigenvectors and corresponding eigenvalues. The eigenvalues represent the variability of the data in that direction. (4) Principal component selection: Based on the magnitude of the eigenvalues, select the top k eigenvectors with the greatest variability as principal components; (5) Dimensionality reduction: Map the original data onto the selected principal components to reduce the data dimensionality; Using principal components extracted by PCA as input features removes redundancy and noise, reduces computational complexity, and improves model prediction accuracy and stability. In addition, the PCA algorithm improves data processing efficiency and enhances model robustness through dimensionality reduction.