Sintered ore production process digital operation guidance method based on data driving

Abstract

The invention relates to the technical field of sinter production condition analysis and production operation guidance, in particular to a sinter production process digital operation guidance method based on data driving, which comprises the following steps of: collecting data in a sintering production process; sorting analysis data, synchronizing the collected modal data to the sinter material in the same range to form a sample to be analyzed; predicting the quality of sintered ore; analyzing a sintering production state; the method comprises the following steps: determining a trigger proposition through a sintered ore prediction quality result and related process parameters, determining a trigger rule through the trigger proposition, determining a to-be-adjusted parameter through the trigger rule, and determining a parameter adjustment amount through historical data comparison calculation to realize sintering production operation guidance output; according to the method, the problems that the sinter production process is complex, the manual adjustment variation is large, and misoperation is prone to occurring are solved, and sinter quality prediction and online production adjustment are comprehensively achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of sinter production status analysis and production operation guidance, and particularly to a data-driven digital operation guidance method for the sinter production process. Background Art

[0002] The sinter production process is a very complex physico-chemical process, and the system has control difficulties such as large lag, multi-variables, time-varying, and some key production parameters being difficult to directly detect online. At present, the production process control management of existing sinter plants is carried out through manual intervention, mainly including the control of the moisture content of the mixed material, the control of the trolley machine speed, the control of the sintering machine igniter, etc. Manual reference to various control parameters and test data is used to decide the actual control scheme, and then it is tested and adjusted during production. This method has problems such as human deviation and misoperation, it is difficult to ensure the quality of sinter, and at the same time it is inefficient and uneconomical, resulting in unnecessary waste. Therefore, the present invention combines multi-field knowledge such as data-driven quality prediction and the rule judgment of an expert system, and proposes a data-driven digital operation guidance method for the sinter production process. Summary of the Invention

[0003] The present invention provides a data-driven digital operation guidance method for the sinter production process, which solves the problems of complex sinter production process, large variation in manual adjustment, and easy occurrence of misoperation, and provides a data-driven digital operation guidance method for the sinter production process, comprehensively realizing the quality prediction of sinter and online production adjustment.

[0004] To achieve the above object, the present invention is implemented by adopting the following technical solutions:

[0005] A data-driven digital operation guidance method for the sinter production process includes the following steps:

[0006] S1. Data collection of the sinter production process;

[0007] S2. Analysis data collation: Synchronize various modal data collected to the same range of sinter materials to form a sample to be analyzed;

[0008] S3. Sinter quality prediction;

[0009] S4. Sinter production status analysis: Determine the trigger proposition through the predicted quality result of sinter and relevant process parameters, determine the trigger rule through the trigger proposition, and determine the parameters to be adjusted and the conclusion parameters through the trigger rule;

[0010] For the parameters to be adjusted (M1, M2, M3... M determined by the trigger rule C), and the conclusion parameters, find the n groups of data (a1, b2, c3,... z1)... (a n , b n , c n ... z n ), and then determine that the adjustment amount of each parameter finally is

[0011]

[0012] where m ∈ (M1, M2, M3... M C ), m is the parameter label; T m is the adjustment amount of the parameter m to be adjusted; V m0 is the current online value of the parameter m to be adjusted, V mi is the i-th matching data of the parameter m to be adjusted; K m is the adjustment coefficient, and K is determined according to the data correlation m is greater than 0 and less than 1; n is the number of groups of matching data;

[0013] S5. Output of sintering production operation guidance: Describe the triggering rules and the recommended adjustment amounts of relevant parameters;

[0014] S6. Steps S1 - S5 are executed in a cycle.

[0015] Furthermore, the sintering production process data includes production process parameters, raw fuel data, environmental data, or online tail section image data and online thermal image data that are relevant to the sinter quality index in the sinter production process.

[0016] Furthermore, the sinter quality prediction is to perform feature vector extraction and feature fusion on the sample to be analyzed, and then conduct in-depth analysis and prediction to obtain the sinter quality index result.

[0017] Furthermore, the sample to be analyzed is the instant data of the monitoring points corresponding to the sinter material running to each production link within the same range.

[0018] Furthermore, the triggering proposition is to use a certain sinter quality parameter or production parameter as a reference item, and divide its numerical change range into multiple regions through different thresholds. If the actual value of this parameter is within a certain region, then this parameter is defined to be in a certain state, and the corresponding proposition is triggered.

[0019] Furthermore, the triggering rule means that both the condition and conclusion parts of the rule are triggered rules. Each condition and conclusion part of the rule corresponds to a proposition. If this proposition is triggered, then this condition and result are triggered.

[0020] Further, the parameter to be adjusted determined according to the trigger rule refers to the sum of the parameters corresponding to the trigger propositions in the condition part of the trigger rule, and the conclusion parameter refers to the sum of the parameters corresponding to the trigger propositions in the conclusion part of the trigger rule.

[0021] Further, the method for finding the n groups of data (a1, b2, c3,... z1) that best match the current online data (a, b, c... z)...

[0022] (a n , b n , c n ... z n ) is as follows: First, for the conclusion parameter, find the target reference data from the historical sample database, that is, determine the optimized parameter status for the conclusion parameter status, further determine the parameter target range, and then find the target reference data that meets all the parameter target range conditions;

[0023] Then, for the further adjusted parameters (M1, M2, M3... M C ) find the N groups of data with the smallest weighted sum of data normalization distance differences from the target reference data;

[0024]

[0025] The above formula is calculated separately for each group of data in the target reference data.

[0026] Compared with the prior art, the beneficial effects of the present invention are:

[0027] 1) It is applicable to the industrial production operation environment, has reliable operation, simple algorithm, and is easy to implement;

[0028] 2) The present invention integrates knowledge in multiple fields such as sintering process and artificial intelligence, makes full use of various information in the sintered ore process, and expands the available information volume;

[0029] 3) Based on data-driven and big data analysis, it effectively overcomes decision-making mistakes caused by knowledge blind spots;

[0030] 4) This method has a high degree of automation and is applicable to various computing devices such as computers and single-chip microcontrollers. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 is the flowchart of the method of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0032] The following further describes the specific embodiments of the present invention with reference to the drawings:

[0033] See Figure 1, which is the flowchart of the method of the present invention. A digital operation guidance method for the sintered ore production process based on data driving according to the present invention comprehensively analyzes the data of the entire chain before the sintered ore production finished product, predicts the sintered ore quality index, and gives a reference adjustment plan through comparison with sample data, including the following steps:

[0034] S1. Sintering production process data acquisition: The sintering production process data obtains the production process parameters, raw fuel data, environmental data, or online tail section image data and online thermal image data that are relevant to the sintered ore quality index in the sintered ore production process; among them, the online tail section image data and online thermal image data can be the images directly collected by a camera or a thermal imager, or the images after splicing the data collected by a camera or a thermal imager, or the images after image preprocessing.

[0035] S2. Analysis data collation: Analysis data collation is to synchronize various modal data to the sintered ore materials in the same range to form a sample to be analyzed; the sample to be analyzed is the instant data of the monitoring points corresponding to each production link when the sintered ore materials in the same range run to, for example, the instant moisture content when the materials pass through the sintered ore mixture moisture detection point, or for example, the ignition temperature when the materials pass through the igniter.

[0036] S3. Sintered ore quality prediction: Sintered ore quality prediction is to perform deep analysis and prediction on the sample to be analyzed after feature vector extraction and feature fusion; among them, feature vector extraction is to represent the original data of different modalities as numerical vectors that can be processed by a computer or further abstracted into higher-level feature vectors; data feature integration can be implemented using a DNN framework, and image feature extraction can be implemented using a CNN framework; feature fusion is to obtain a fused feature vector according to a certain rule for various single-modal feature representations; there are various feature fusion methods, such as simple operations without introducing parameters like weighted linear splicing, tensor Cartesian product multiplication, etc., and fully connected networks that automatically learn correlations; deep analysis and prediction is to further perform deep learning on the fused feature data to obtain the results of relevant sintered ore quality indexes, and the deep learning framework can adopt an RNN model.

[0037] S4. Sintering production status analysis: Determine the trigger proposition through the predicted quality results of the sintered ore and relevant process parameters, determine the trigger rule through the trigger proposition, determine the parameter to be adjusted through the trigger rule, and then determine the parameter adjustment amount through comparison and calculation with historical data;

[0038] The triggering proposition is aimed at a certain sinter quality parameter or production parameter as a reference item. Its numerical range is divided into multiple states through different thresholds, such as too high, moderate, too low, too fast. If the actual value of the parameter is within a certain range, it is defined that the parameter is in a certain state, and the corresponding proposition is triggered. For example, the proposition is similar to "The ignition temperature is less than 700 degrees, and the ignition temperature is too low". When the ignition temperature is less than 700 degrees, this proposition is triggered;

[0039] The triggering rule means that both the condition and conclusion parts of the rule are triggered rules. Each condition and conclusion part of the rule corresponds to a proposition. If the proposition is triggered, then the condition and conclusion are triggered. The rule is determined by process theory. For example, "If the ignition temperature is too low and the sinter permeability is poor, then the FeO content may be high". Comparing with the actual situation, if the actual ignition temperature is less than 700 degrees, the proposition related to the too low ignition temperature is triggered. If the pressure of multiple detection air boxes is large and the main pipe negative pressure is large, the proposition related to the poor sinter permeability is triggered. At the same time, if the predicted FeO content is greater than 9, the proposition related to the high FeO content is triggered, then the above rule is triggered;

[0040] The parameter to be adjusted determined by the triggering rule refers to the sum of the parameters corresponding to the triggering propositions in the condition part of the triggering rule. The conclusion parameter refers to the sum of the parameters corresponding to the triggering propositions in the conclusion part of the triggering rule. For example, a triggering rule "If the ignition temperature is too low and the sinter permeability is poor, then the FeO content may be high", the corresponding parameter to be adjusted is the ignition temperature, and the conclusion parameter is the FeO content. At the same time, there may be multiple triggering rules. The parameter to be adjusted is the sum of the proposition parameters corresponding to the condition parts of multiple rules, and the conclusion parameter is the sum of the proposition parameters corresponding to the conclusion parts of multiple rules;

[0041] For the parameter to be adjusted (M1, M2, M3... M C ) and the conclusion parameter determined by the triggering rule, find the n groups of data (a1, b2, c3,... z1)... (a n , b n , c n ... z n ) that are most matched with the current online data (a, b, c... z) from the historical sample database;

[0042] The search method is as follows: First, for the conclusion parameter, find the target reference data from the historical sample database. That is, for the conclusion parameter status, such as high FeO content and low conversion index, determine the optimized parameter status, such as moderate FeO content and moderate conversion index. Further determine the parameter target range, that is, the parameter range corresponding to the optimized parameter status. For example, the FeO content is greater than 7% and less than 9%, and the conversion index is greater than 80.5 and less than 81.5. Then find the target reference data group that meets all the parameter target range conditions. Then, for the parameters to be adjusted (M1, M2, M3... M C ) find N groups of data with the smallest weighted sum of data normalization distance differences from the target reference data;

[0043]

[0044] where m ∈ (M1, M2, M3... M C ), m is the parameter label; T m is the adjustment amount of the parameter m to be adjusted; V m0 is the current online value of the parameter m to be adjusted, V mi is the i-th matching data of the parameter m to be adjusted; K m is the adjustment coefficient, and K is determined according to the data correlation m is greater than 0 and less than 1; n is the number of groups of matching data.

[0045] S5. Output of sintering production operation guidance: Describe the trigger rule and the recommended adjustment amount of relevant parameters.

[0046] S6. Steps S1 - S5 are executed cyclically according to a period.

[0047] The above embodiments are implemented on the premise of the technical solution of the present invention, and the detailed implementation manners and specific operation processes are given. However, the protection scope of the present invention is not limited to the above embodiments. The methods used in the above embodiments are all conventional methods unless otherwise specified.

Claims

1. A digital operation guidance method for the sinter production process based on data-driven, characterized in that, It includes the following steps: S1. Collecting sintering production process data; S2. Analyzing data arrangement: Synchronizing various modal data collected to the sinter ore materials within the same range to form a sample to be analyzed; S3. Predicting the quality of sinter ore; S4. Analyzing the sintering production status: Determining the trigger proposition through the predicted quality result of sinter ore and relevant process parameters, determining the trigger rule through the trigger proposition, and determining the parameters to be adjusted and the conclusion parameters through the trigger rule; For the parameters to be adjusted (M1, M2, M3... M C ) and the conclusion parameters determined by the trigger rule, find the n groups of data (a1, b2, c3,... z1)... (a n , b n , c n ... z n ) that are most matched with the current online data (a, b, c... z) from the historical sample database, and then determine that the adjustment amount of each parameter finally is Among them, m ∈ (M1, M2, M3... M C ), where m is a parameter label; T m is the adjustment amount of the parameter m to be adjusted; V m0 is the current online value of the parameter m to be adjusted, V mi is the i-th matching data of the parameter m to be adjusted; K m is an adjustment coefficient, determined according to the data correlation, K m is greater than 0 and less than 1; n is the number of groups of matching data; S5. Outputting sintering production operation guidance: Describing the trigger rule and the recommended adjustment amount of relevant parameters; S6. Steps S1 - S5 are executed cyclically according to a period.

2. The digital operation guidance method for the sinter production process based on data driving according to claim 1, characterized in that, The sintering production process data includes production process parameters, raw fuel data, environmental data, or online tail section image data and online thermal image data that are relevant to the sinter ore quality index in the sinter ore production process.

3. A digital operation guidance method for the sinter production process based on data driving according to claim 1, characterized in that, The prediction of the sinter ore quality is to perform deep analysis and prediction after extracting feature vectors and fusing features from the sample to be analyzed to obtain the sinter ore quality index result.

4. A digital operation guidance method for the sinter production process based on data driving according to claim 1, characterized in that, The sample to be analyzed is the real-time data of the sinter ore materials within the same range at each monitoring point corresponding to each production link.

5. A digital operation guidance method for the sinter production process based on data driving according to claim 1, characterized in that The trigger proposition is to take a certain sinter ore quality parameter or production parameter as a reference item, divide its numerical change range into multiple regions through different thresholds. If the actual value of this parameter is within a certain region, then it is defined that this parameter is in a certain state, and the corresponding proposition is triggered.

6. The digital operation guidance method for the sinter production process based on data driving according to claim 1, wherein, The trigger rule means that both the condition and conclusion parts of the rule are triggered rules. Each condition and conclusion part of the rule corresponds to a proposition. If this proposition is triggered, then this condition and result are triggered.

7. A digital operation guidance method for the sinter production process based on data driving according to claim 1, characterized in that The parameter to be adjusted determined for the trigger rule refers to the sum of the parameters corresponding to the trigger proposition corresponding to the condition part of the trigger rule, and the conclusion parameter refers to the sum of the parameters corresponding to the trigger proposition corresponding to the conclusion part of the trigger rule.

8. A digital operation guidance method for the sinter production process based on data driving according to claim 1, characterized in that, The method of finding the n groups of data (a1, b2, c3,... z1)... (a n , b n , c n ... z n ) that best match the current online data (a, b, c... z) is as follows: First, for the conclusion parameters, find the target reference data from the historical sample database. That is, for the conclusion parameter status, determine the optimized parameter status, further determine the parameter target range, and then find the target reference data that meets all the parameter target range conditions; Then, for further adjusting the parameters (M1, M2, M3... M C ), find N groups of data with the smallest weighted sum of data normalization distance differences from the target reference data; The above formula is calculated separately for each group of data in the target reference data.

Images