Method for improving continuous casting stability of low-silicon aluminum-containing steel

Abstract

The present application belongs to the technical field of metallurgical continuous casting process, and provides a method for improving the continuous casting stability of low-silicon aluminum-containing steel, which comprises constructing a first prediction model and a first prediction model for predicting the mold change time based on the protective slag content and the liquid slag thickness respectively through an LSTM model, taking the minimum value of the model output result as the target period, analyzing the fluctuation of the target period sequence, adjusting the mean value when it is stable, determining the optimal period through clustering when it is unstable, simplifying the analysis process if the mold change operation is concentrated on a single factor, reducing redundant judgments, realizing accurate regulation and control of the mold change period through multi-parameter modeling and intelligent algorithms, reducing the cost of manual intervention, improving the continuous casting efficiency and equipment life, and having significant economic benefits.

Description

Technical Field

[0001] This invention belongs to the field of metallurgical continuous casting technology, specifically a method and system for improving the casting stability of low-silicon aluminum-containing steel. Background Technology

[0002] With the development of high-alloy building steel smelting, how to achieve stable continuous casting based on high-alloy building steel smelting has become a research hotspot in the field of steel metallurgical process innovation. However, there are still significant technical bottlenecks in the continuous casting process. Therefore, it is necessary to develop a method to improve the stability of continuous casting of low-silicon aluminum steel.

[0003] In the continuous casting process, the protective slag is a key auxiliary material, and its performance directly affects the lubrication, heat preservation, and billet quality of the molten steel in the crystallizer. However, the existing control methods for slag replacement have many limitations and cannot meet the needs of efficient and stable continuous casting production. On the one hand, the judgment of the timing of slag replacement is one-sided and lagging. Existing technologies mostly rely on a single indicator to make slag replacement decisions without considering the synergistic effect of multiple factors, resulting in slag replacement being too early or too late. Furthermore, the failure of the protective slag performance can cause surface defects in the billet, affecting the stability of continuous casting.

[0004] On the other hand, the lack of a multi-factor collaborative decision-making mechanism means that when the composition content of the protective slag and the thickness of the liquid slag are used to determine the protective slag replacement operation, the existing technology has not established a collaborative judgment logic between the two. If the timing of the two judgment and slag replacement operations conflicts, there is a lack of scientific selection criteria, which may lead to billet quality problems or production interruptions, affecting the stability of continuous casting.

[0005] Therefore, the present invention provides a method for improving the casting stability of low-silicon aluminum-containing steel. Summary of the Invention

[0006] In order to overcome the shortcomings of the prior art, at least one technical problem raised in the background art is solved.

[0007] The technical solution adopted by this invention to solve its technical problem is: a method for improving the casting stability of low-silicon aluminum-containing steel, comprising:

[0008] Based on the Al2O3 content of the protective slag, determine whether a slag replacement operation caused by Al2O3 content is required. If not, establish a first slag replacement prediction model to predict the first slag replacement time.

[0009] If the slag replacement operation caused by Al2O3 content is not required, determine whether the slag replacement operation caused by liquid slag thickness is required. If not, establish a second slag replacement prediction model, predict the second slag replacement time, and determine the target slag replacement cycle.

[0010] Based on the target slag replacement cycle, a target cycle sequence is constructed and stability analysis is performed to determine the adjustable slag replacement cycle and whether the initial slag replacement cycle needs to be adjusted. If so, the initial slag replacement cycle is adjusted.

[0011] A volatility analysis is performed on the initial slag replacement cycle. When the volatility is large, it is determined whether there is a concentration phenomenon in the initial slag replacement cycle. If so, the concentration object of the initial slag replacement cycle is identified, and the adjustment method of the initial slag replacement cycle is optimized based on the concentration object.

[0012] Furthermore, the process of establishing the first slag replacement prediction model is as follows:

[0013] The continuous casting speed V, the temperature T when the molten steel enters the crystallizer, the Al2O3 formation rate C, and the Al2O3 content were obtained at the same time point during multiple historical continuous casting stages. Standard Al2O3 content and V, T, C at the same time point Record as a group;

[0014] A first slag replacement prediction model was constructed based on an LSTM neural network.

[0015] Input layer: A set of V, T, C, As an input item, and input multiple times;

[0016] Hidden layer: By capturing the weighting coefficients between V, T, and C through the interaction of gating mechanisms and cell state, a weighting formula is established to calculate the rate of change of Al2O3 content over time;

[0017] Output layer: computation and The difference between the values ​​will be compared with the rate of change of Al2O3 content over time to obtain the predicted time when the Al2O3 content at the current time point reaches the standard Al2O3 content, and then output.

[0018] Furthermore, the process of predicting the first slag replacement time is as follows:

[0019] Obtain the Al2O3 content of the protective slag at the current time point. Continuous casting speed V, temperature T, standard Al2O3 content

[0020] Obtain the average Al2O3 formation rate C for each historical slag replacement cycle;

[0021] The current time point of the protective slag's V, T, C, Input the data into the first slag replacement prediction model and output the predicted first slag replacement time.

[0022] The model construction and prediction process for predicting the second slag replacement time are the same as those for predicting the first slag replacement time.

[0023] Furthermore, the process of determining the target slag replacement cycle is as follows:

[0024] Get the current time point and the end time point of the last slag replacement, and perform difference processing to obtain the working time of the protective slag at the current time point;

[0025] The working time of the protective slag is added to the first slag replacement time and the second slag replacement time respectively to obtain the first slag replacement cycle and the second slag replacement cycle.

[0026] The first and second slag replacement cycles are compared, and the minimum value is taken as the target slag replacement cycle.

[0027] Furthermore, the process of constructing the target periodic sequence and performing stability analysis is as follows:

[0028] The target slag replacement cycle is obtained multiple times, and the target slag replacement cycle is arranged in time sequence to obtain the target cycle sequence;

[0029] Variance processing is performed on all target slag replacement cycles within the target cycle sequence to obtain the cycle stability coefficient. The cycle stability coefficient is then compared with a cycle stability coefficient threshold.

[0030] When the periodic stability coefficient is less than or equal to the periodic stability coefficient threshold, the target periodic sequence is considered stable.

[0031] Conversely, the target periodic sequence is judged to be unstable.

[0032] Furthermore, the process of determining the adjustable slag replacement cycle is as follows:

[0033] If the target periodic sequence is stable, calculate the average period of all target slag replacement periods within the target periodic sequence and mark it as the adjustable slag replacement period;

[0034] If the target periodic sequence is unstable, cluster analysis is performed on the target periodic sequence to obtain the adjustable slag period.

[0035] Further, determine whether the initial slag replacement cycle needs to be adjusted. If so, the process for adjusting the initial slag replacement cycle is as follows:

[0036] Compare the adjustable slag replacement cycle with the initial slag replacement cycle:

[0037] If the adjustable slag replacement cycle is greater than or equal to the initial slag replacement cycle, it is determined that no adjustment to the initial slag replacement cycle is required.

[0038] Conversely, it is determined that the initial slag replacement cycle needs to be adjusted;

[0039] When the initial slag replacement cycle needs to be adjusted, the initial slag replacement cycle will be updated and adjusted to be consistent with the adjustable slag replacement cycle.

[0040] Furthermore, the process of adjusting the volatility of the initial slag replacement cycle is as follows:

[0041] During the historical slag replacement process, obtain the adjustable slag replacement cycle each time the initial slag replacement cycle is updated and adjusted.

[0042] The cycle durations corresponding to the adjustable slag replacement cycle are integrated into a time-series cycle duration set according to the time sequence corresponding to the adjustable slag replacement cycle.

[0043] Calculate the coefficient of variation of the time series period duration set to obtain the adjusted period fluctuation value;

[0044] The adjustment cycle fluctuation value is compared with the adjustment cycle fluctuation threshold.

[0045] If the adjustment cycle fluctuation value is greater than or equal to the adjustment cycle fluctuation threshold, it is determined that the adjustment volatility of the initial slag replacement cycle is large.

[0046] Conversely, if the initial slag replacement cycle shows little fluctuation, it is determined that the adjustment volatility is small.

[0047] Furthermore, determine whether there is a concentration of adjustments during the initial slag replacement cycle. If so, the process for identifying the concentration object is as follows:

[0048] Obtain the adjustable slag replacement cycle based on the initial slag replacement cycle when making cycle adjustments, and obtain the target cycle sequence corresponding to the adjustable slag replacement cycle. Summarize all target cycle sequences according to time sequence to obtain a mixed cycle sequence.

[0049] Obtain the number of the first slag replacement cycle and the second slag replacement cycle in the mixed cycle sequence, and denot them as the first adjustment cycle number and the second adjustment cycle number;

[0050] The first and second adjustment period numbers are respectively compared with the total number of periods of the mixed periodic sequence to obtain the first set of discriminant values ​​and the second set of discriminant values.

[0051] The discriminant values ​​in the first and second ensembles are compared with the ensemble thresholds, respectively.

[0052] If the first set of discrimination values ​​is greater than or equal to the set discrimination threshold, it indicates that there is an adjustment set phenomenon in the initial slag replacement cycle, and the set object is the first slag replacement cycle.

[0053] If the second set of discrimination values ​​is greater than or equal to the set of discrimination threshold, it indicates that there is an adjustment set of values ​​in the initial slag replacement cycle, and the set of values ​​is the second slag replacement cycle.

[0054] All other comparison cases were determined to be free of adjustment concentration phenomena in the initial slag replacement cycle.

[0055] Furthermore, the process of optimizing the adjustment method of the initial slag replacement cycle based on the adjustment focus object is as follows:

[0056] If the focus is on the first slag replacement cycle, then in the subsequent process of determining the adjustable slag replacement cycle, only the first slag replacement time and the working time of the protective slag are summed, and the summation result is used as the target slag replacement cycle. Through stability analysis, the adjustable slag replacement cycle is determined, and the initial slag replacement cycle is updated and adjusted to the adjustable slag replacement cycle.

[0057] If the focus is on the second slag replacement cycle, then in the subsequent process of determining the adjustable slag replacement cycle, only the second slag replacement time and the working time of the protective slag are summed, and the summation result is used as the target slag replacement cycle. Through stability analysis, the adjustable slag replacement cycle is determined, and the initial slag replacement cycle is updated and adjusted to the adjustable slag replacement cycle.

[0058] The beneficial effects of this invention are as follows:

[0059] Effect 1: Improve the accuracy of slag replacement prediction. By constructing a first prediction model and a second prediction model, the minimum value of the two prediction times is taken as the target slag replacement cycle. This avoids the limitations of single-factor prediction, realizes quantitative prediction of slag replacement time, and improves prediction accuracy.

[0060] Effect 2: When adjusting the slag replacement cycle based on the Al2O3 content and liquid slag thickness of the protective slag, if there is a conflict in the adjustment logic between the two, the adjustment method will be optimized according to the collaborative judgment logic to avoid billet quality problems or production interruptions, thereby improving the stability of continuous casting. Attached Figure Description

[0061] The invention will now be further described with reference to the accompanying drawings.

[0062] Figure 1 This is a schematic diagram of the steps in a method for improving the casting stability of low-silicon aluminum-containing steel according to an embodiment of the present invention;

[0063] Figure 2 This is a schematic diagram of the logic judgment of a method for improving the stability of continuous casting of low-silicon aluminum steel according to an embodiment of the present invention. Detailed Implementation

[0064] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention will be further described below in conjunction with specific embodiments.

[0065] Please see Figure 1 - Figure 2As shown in the embodiment of the present invention, a method for improving the casting stability of low-silicon aluminum-containing steel includes the following steps:

[0066] Step 1: Determine whether a slag replacement operation is needed based on the Al2O3 content of the protective slag. If not, establish a first slag replacement prediction model based on the Al2O3 content data of the protective slag during the historical continuous casting stage to predict the first slag replacement time.

[0067] In step one, the process of determining whether a slag replacement operation is needed due to Al2O3 content is as follows:

[0068] The composition of the protective slag at the current time point during the continuous casting stage was analyzed using X-ray fluorescence spectroscopy to obtain the Al2O3 content of the protective slag.

[0069] The Al2O3 content of the protective flux was compared with the standard Al2O3 content of the protective flux. Comparison:

[0070] Standard Al2O3 content and standard liquid slag thickness H c2 The settings are determined by the staff based on the actual working conditions;

[0071] If the Al2O3 content of the protective slag is greater than or equal to the standard Al2O3 content of the protective slag, it is determined that a slag replacement operation is required.

[0072] If the Al2O3 content of the protective slag is less than the standard Al2O3 content of the protective slag, it is determined that no slag replacement operation is required.

[0073] In step one, the process of establishing the first slag replacement prediction model is as follows:

[0074] The continuous casting speed V, the temperature T when the molten steel enters the crystallizer, the Al2O3 formation rate C, and the Al2O3 content were obtained at the same time point during the historical continuous casting stages multiple times. Standard Al2O3 content and V, T, C at the same time point Record as a group;

[0075] A first slag replacement prediction model was constructed based on an LSTM neural network.

[0076] Input layer: A set of V, T, C, As an input item, and input multiple times;

[0077] Hidden layer: By capturing the weighting coefficients between V, T, and C through the interaction of gating mechanisms and cell state, a weighting formula is established to calculate the rate of change of Al2O3 content over time;

[0078] Output layer: computation and The difference between the two values ​​is calculated by ratioing the difference to the rate of change of Al2O3 content over time, thus obtaining the Al2O3 content at the current time point that meets the standard Al2O3 content. The predicted time is then output.

[0079] In step one, the process of predicting the first slag replacement time is as follows:

[0080] The continuous casting speed V at the current continuous casting stage is read using a programmable logic controller (PLC).

[0081] An infrared thermometer is installed above the crystallizer inlet to obtain the temperature T at the current continuous casting stage.

[0082] Obtain the average Al2O3 formation rate C for each historical slag replacement cycle;

[0083] Standard Al2O3 content

[0084] V, T, C in the current continuous casting stage Input the first slag replacement prediction model and output the first slag replacement time;

[0085] Step 2: If the slag replacement operation caused by Al2O3 content is not required, determine whether the slag replacement operation caused by the slag thickness is necessary based on the slag thickness data. If not, establish a second slag replacement prediction model to predict the second slag replacement time, and determine the target slag replacement cycle by comparing it with the first slag replacement time.

[0086] In step two, the process of determining whether a slag replacement operation is necessary due to the thickness of the liquid slag is as follows:

[0087] The thickness H of the liquid slag at the current time point is obtained using an X-ray thickness sensor. c实 ;

[0088] Compare the current liquid sludge thickness with the standard liquid sludge thickness:

[0089] If the thickness of the liquid sludge at the current time point is greater than or equal to the standard thickness of the liquid sludge, it is determined that a sludge replacement operation is required.

[0090] If the thickness of the liquid sludge at the current time point is less than the standard thickness of the liquid sludge, then it is determined that no sludge replacement operation is required;

[0091] In step two, the process of establishing the second slag replacement prediction model is as follows:

[0092] The melting point M of the protective slag at the same time point during the historical continuous casting stage was obtained multiple times. p Viscosity V of protective slag sCrystallizer vibration frequency F, liquid slag thickness H c历 Standard liquid slag thickness H c2 ;

[0093] A second slag replacement prediction model was constructed based on an LSTM neural network.

[0094] Input layer: The melting point M of the protective slag at the same time point is repeatedly set. p Viscosity V of protective slag s Crystallizer vibration frequency F, liquid slag thickness H c历 Standard liquid slag thickness H c2 Use it as input features;

[0095] Hidden layers: Capture H through gating mechanisms and interactions with cell states. c历 With M p V s The weighting coefficients between F and H are established, and a weighting formula is constructed to calculate H. c历 Rate of change over time;

[0096] Output layer: Calculate H c历 With H c2 The difference between them, and H c历 The ratio of the rate of change over time is used to obtain H at the current time point. c历 Reaching H c2 The predicted time is then output.

[0097] In step two, the process of predicting the second slag replacement time is as follows:

[0098] Obtain the current melting point M of the protective slag from the instruction manual. p1 Viscosity V of protective slag s1 ;

[0099] The current crystallizer vibration frequency F1 is read through the crystallizer control platform;

[0100] Standard liquid slag thickness H c1 ;

[0101] M in the current continuous casting stage p1 V s1 F1, H c实 H c1 Input the data into the second slag replacement prediction model and output the second slag replacement time;

[0102] In step two, the process of determining the target slag replacement cycle is as follows:

[0103] The current time point and the end time point of the previous slag replacement are obtained by a programmable logic controller (PLC), and the difference is processed to obtain the working time of the protective slag at the current time point.

[0104] The working time of the protective slag is added to the first slag replacement time and the second slag replacement time respectively to obtain the first slag replacement cycle and the second slag replacement cycle.

[0105] Compare the first and second slag replacement cycles and take the minimum value as the target slag replacement cycle.

[0106] It should be noted that the reason for taking the minimum value as the target slag replacement cycle is as follows:

[0107] If both the first and second slag replacement cycles need to be considered, and either of the influencing factors causes the protective slag to need to be replaced, a slag replacement operation is required. Therefore, taking the minimum value can ensure that one side performs the slag replacement operation at the predicted time point to meet the slag replacement requirements, while the other side will also perform the slag replacement operation at the predicted time point to avoid subsequent problems caused by the failure to perform the slag replacement operation in time.

[0108] Step 3: Construct the target cycle sequence and perform stability analysis. Determine the adjustable slag replacement cycle based on the stability analysis results and compare it with the initial slag replacement cycle to determine whether the initial slag replacement cycle needs to be adjusted. If so, adjust the initial slag replacement cycle.

[0109] In step three, the process of constructing the target periodic sequence and performing stability analysis is as follows:

[0110] The target slag replacement cycle is obtained multiple times, and the target slag replacement cycle is arranged in time sequence to obtain the target cycle sequence;

[0111] The target periodic sequence is subjected to variance processing to obtain the periodic stability coefficient, which is then compared with a periodic stability coefficient threshold.

[0112] The threshold for the periodic stability coefficient is set by the worker based on the actual situation;

[0113] When the periodic stability coefficient is less than or equal to the periodic stability coefficient threshold, the target periodic sequence is considered stable.

[0114] When the periodic stability coefficient is greater than the periodic stability coefficient threshold, the target periodic sequence is determined to be unstable.

[0115] In step three, the process of determining the adjustable slag replacement cycle based on the stability analysis results is as follows:

[0116] If the target periodic sequence is stable, calculate the average period of all target slag replacement periods within the target periodic sequence and mark it as the adjustable slag replacement period;

[0117] If the target periodic sequence is unstable, cluster analysis is performed on the target periodic sequence to obtain the adjustable slag period;

[0118] The analysis process for clustering the target periodic sequence is as follows:

[0119] A1: Obtain the feature values ​​of the target periodic sequence, including but not limited to the average period, period variance, maximum period and minimum period of all target slag replacement periods in the target periodic sequence;

[0120] A2: Set the data radius (eps) and minimum number of sample points (min_samples):

[0121] The minimum number of sample points is usually set to the data dimension + 1 (e.g., 3 for two-dimensional data);

[0122] The data radius uses a k-distance curve to calculate the distance from each point to its min_samples nearest neighbor, sorts the data by distance and plots the curve, and selects the inflection point where the curve rises significantly as the eps value.

[0123] A3: Input the feature values ​​of the target cycle sequence into the DBSCAN clustering algorithm. By automatically adjusting the data radius and the minimum number of sample points, each slag replacement cycle is assigned to a cluster.

[0124] A4: Plot typical periodic curves within a cluster (e.g., the period of one cluster is concentrated in 4-5 hours, while that of another cluster is in 8-10 hours);

[0125] A5: Count the number of cycles in each cluster, select the cluster with the most cycles as the target cluster, and average the cycles within the cluster to obtain the adjustable slag cycle;

[0126] It should be noted that the reason for selecting the adjustable slag replacement cycle is as follows:

[0127] When the mean of the target cycle sequence is less than the existing slag replacement cycle, the accuracy of the mean of the target cycle sequence is small, indicating that the actual slag replacement cycle cannot meet the current slag replacement operation and is likely to affect the continuous casting process effect. When the mean of the target cycle sequence is less than the existing slag replacement cycle, the accuracy of the existing slag replacement cycle is small, and it can meet the current work needs. Therefore, there is no need to adjust the slag replacement cycle, thus reducing process costs.

[0128] In step three, it is determined whether the initial slag replacement cycle needs to be adjusted. If so, the process for adjusting the initial slag replacement cycle is as follows:

[0129] Comparative analysis of the adjustable slag replacement cycle and the initial slag replacement cycle:

[0130] If the adjustable slag replacement cycle is greater than or equal to the initial slag replacement cycle, it is determined that no adjustment to the initial slag replacement cycle is required.

[0131] If the adjustable slag replacement cycle is less than the initial slag replacement cycle, it is determined that the initial slag replacement cycle needs to be adjusted.

[0132] When it is necessary to adjust the initial slag replacement cycle, the initial slag replacement cycle will be updated and adjusted to be consistent with the adjustable slag replacement cycle.

[0133] Step 4: Analyze the adjustment volatility of the initial slag replacement cycle, and when the adjustment volatility is large, determine whether there is a concentration of adjustments in the initial slag replacement cycle. If so, identify the adjustment focus of the initial slag replacement cycle and optimize the adjustment method of the initial slag replacement cycle based on the adjustment focus.

[0134] In step four, the process of analyzing the adjustment volatility of the initial slag replacement cycle is as follows:

[0135] During the historical slag replacement process, obtain the adjustable slag replacement cycle each time the initial slag replacement cycle is updated and adjusted.

[0136] It should be noted that the initial slag replacement cycle is adjusted according to the adjustable slag replacement cycle;

[0137] The cycle durations corresponding to the adjustable slag replacement cycle are integrated into a time-series cycle duration set according to the time sequence corresponding to the adjustable slag replacement cycle.

[0138] Calculate the coefficient of variation of the time series period duration set to obtain the adjusted period fluctuation value;

[0139] The adjustment cycle fluctuation value is compared with the adjustment cycle fluctuation threshold.

[0140] The threshold for adjusting periodic fluctuations is set by experts;

[0141] If the adjustment cycle fluctuation value is greater than or equal to the adjustment cycle fluctuation threshold, it is determined that the adjustment volatility of the initial slag replacement cycle is large.

[0142] If the adjustment cycle fluctuation value is less than the adjustment cycle fluctuation threshold, it is determined that the adjustment fluctuation of the initial slag replacement cycle is small.

[0143] In step four, it is determined whether there is a concentration of adjustments in the initial slag replacement cycle. If so, the process for identifying the concentration object is as follows:

[0144] Obtain the adjustable slag replacement cycle based on the initial slag replacement cycle when making cycle adjustments, and obtain the target cycle sequence corresponding to the adjustable slag replacement cycle. Summarize all target cycle sequences according to time sequence to obtain a mixed cycle sequence.

[0145] The target cycle sequence contains multiple first slag replacement cycles and second slag replacement cycles, and the mixed cycle sequence also contains multiple first slag replacement cycles and second slag replacement cycles respectively.

[0146] Obtain the number of the first slag replacement cycle and the second slag replacement cycle in the mixed cycle sequence, and denot them as the first adjustment cycle number and the second adjustment cycle number;

[0147] The first and second adjustment period numbers are respectively compared with the total number of periods of the mixed periodic sequence to obtain the first set of discriminant values ​​and the second set of discriminant values.

[0148] Wherein, the total number of periods in the mixed periodic sequence is the total number of periods in the mixed periodic sequence;

[0149] It should be noted that the physical significance of obtaining the discriminant values ​​from the first and second episodes lies in:

[0150] The adjustment settling discriminant value is obtained by comparing the first adjustment cycle number and the second adjustment cycle number with the total number of cycles in the mixed cycle sequence. The larger the first settling discriminant value, the more first slag replacement cycles there are, which means the first slag replacement cycle accounts for a higher proportion. The larger the second settling discriminant value, the more second slag replacement cycles there are, which means the second slag replacement cycle accounts for a higher proportion.

[0151] The discriminant values ​​in the first and second ensembles are compared with the ensemble thresholds, respectively.

[0152] Among them, the centralized discrimination threshold is set by staff based on historical actual conditions;

[0153] If the first set of discrimination values ​​is greater than or equal to the set discrimination threshold, it indicates that there is an adjustment set phenomenon in the initial slag replacement cycle, and the set object is the first slag replacement cycle.

[0154] If the second set of discrimination values ​​is greater than or equal to the set of discrimination threshold, it indicates that there is an adjustment set of values ​​in the initial slag replacement cycle, and the set of values ​​is the second slag replacement cycle.

[0155] All other comparison cases were determined to indicate that there was no concentrated adjustment phenomenon in the initial slag replacement cycle;

[0156] In step four, the process of optimizing the adjustment method of the initial slag replacement cycle based on the adjustment focus is as follows:

[0157] If the focus is on the first slag replacement cycle, then in the subsequent process of determining the adjustable slag replacement cycle, only the first slag replacement time and the working time of the protective slag are summed, and the summation result is used as the target slag replacement cycle. Through stability analysis, the adjustable slag replacement cycle is determined, and the initial slag replacement cycle is updated and adjusted to the adjustable slag replacement cycle.

[0158] If the focus is on the second slag replacement cycle, then in the subsequent process of determining the adjustable slag replacement cycle, only the second slag replacement time and the working time of the protective slag will be summed, and the summed result will be used as the target slag replacement cycle. Through stability analysis, the adjustable slag replacement cycle will be determined, and the initial slag replacement cycle will be updated and adjusted to the adjustable slag replacement cycle.

[0159] The technical solution of this invention is as follows: Based on the Al2O3 content of the protective slag, it is determined whether a slag replacement operation caused by the Al2O3 content is required. If not, a first slag replacement prediction model is established to predict the first slag replacement time. If a slag replacement operation caused by the Al2O3 content is not required, it is determined whether a slag replacement operation caused by the liquid slag thickness is required. If not, a second slag replacement prediction model is established to predict the second slag replacement time, and a target slag replacement cycle is determined. Based on the target slag replacement cycle, a target cycle sequence is constructed and stability analysis is performed to determine the adjustable slag replacement cycle. It is then determined whether the initial slag replacement cycle needs to be adjusted. If so, the initial slag replacement cycle is adjusted, and the adjustment volatility of the initial slag replacement cycle is analyzed. When the adjustment volatility is large, it is determined whether there is an adjustment concentration phenomenon in the initial slag replacement cycle. If so, the adjustment concentration object of the initial slag replacement cycle is determined, and the adjustment method of the initial slag replacement cycle is optimized.

[0160] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for improving the casting stability of low-silicon aluminum-containing steel, characterized in that: include: Determine whether it is necessary to perform If the slag replacement operation caused by the content is not required, establish a first slag replacement prediction model to predict the first slag replacement time. The process of establishing the first slag replacement prediction model is as follows: The continuous casting speed V and the temperature T of the molten steel entering the crystallizer were obtained multiple times at the same time point during the historical continuous casting stages. Generation rate C, Content, Standard content and at the same time point Record as a group; A first slag replacement prediction model was constructed based on an LSTM neural network. Input layer: As an input item, and input multiple times; Hidden layers: Capture [data / capture] through gating mechanisms and interactions with cell states. We establish a weighting formula based on the weighting coefficients between them and calculate the weighting coefficients. Rate of change of content over time; Output layer: computation and The difference between them will be with The ratio of the rate of change of content over time is used to obtain the value at the current time point. Content meets the standard The prediction time of the content is calculated and output. The process of predicting the first replacement time is as follows: Get the protective residue at the current time point Content ȵ, continuous casting speed V, temperature T, standard content ; Obtain the average value for each historical slag replacement cycle. Generation rate C; Protect the slag at the current time point Input the data into the first slag replacement prediction model and output the predicted first slag replacement time. Determine whether a slag replacement operation is needed due to the thickness of the liquid slag. If not, establish a second slag replacement prediction model to predict the second slag replacement time and determine the target slag replacement cycle. The process of establishing the second slag replacement prediction model is as follows: Multiple acquisitions of the melting point of the protective slag at the same time point during the historical continuous casting stage Viscosity of protective slag Crystallizer vibration frequency F, liquid slag thickness Standard liquid slag thickness ; A second slag replacement prediction model was constructed based on an LSTM neural network. Input layer: The melting point of the protective slag at the same time point is repeatedly measured. Viscosity of protective slag Crystallizer vibration frequency F, liquid slag thickness Standard liquid slag thickness Use it as input features; Hidden layers: Capture [data / capture] through gating mechanisms and interactions with cell states. The weighting coefficients between them are determined, and a weighting formula is established to calculate the weighting coefficients. Rate of change over time; Output layer: computation and The difference between them, and the difference and The ratio of the rate of change over time is used to obtain the current time point. achieve The predicted time is then output. The process of predicting the second slag replacement time is as follows: Obtain the current melting point of the protective slag from the instruction manual. Viscosity of protective slag ; The current crystallizer vibration frequency is read through the crystallizer control platform. ; Standard liquid slag thickness ; The current continuous casting stage Input the data into the second slag replacement prediction model and output the second slag replacement time; The process of determining the target slag replacement cycle is as follows: Get the current time point and the end time point of the last slag replacement, and perform difference processing to obtain the working time of the protective slag at the current time point; The working time of the protective slag is added to the first slag replacement time and the second slag replacement time respectively to obtain the first slag replacement cycle and the second slag replacement cycle. Compare the first and second slag replacement cycles and take the minimum value as the target slag replacement cycle. Construct the target cycle sequence and perform stability analysis to determine the adjustable slag replacement cycle and whether the initial slag replacement cycle needs to be adjusted. If so, adjust the initial slag replacement cycle. A volatility analysis is performed on the initial slag replacement cycle. If the volatility is large, it is determined whether there is a concentration of adjustments in the initial slag replacement cycle. If so, the adjustment focus of the initial slag replacement cycle is identified, and the adjustment method of the initial slag replacement cycle is optimized based on the adjustment focus.

2. The method for improving the casting stability of low-silicon aluminum-containing steel according to claim 1, characterized in that: The process of constructing the target periodic sequence and performing stability analysis is as follows: The target slag replacement cycle is obtained multiple times, and the target slag replacement cycle is arranged in time sequence to obtain the target cycle sequence; Variance processing is performed on all target slag replacement cycles within the target cycle sequence to obtain the cycle stability coefficient. The cycle stability coefficient is then compared with a cycle stability coefficient threshold. When the periodic stability coefficient is less than or equal to the periodic stability coefficient threshold, the target periodic sequence is considered stable. Conversely, the target periodic sequence is judged to be unstable.

3. The method for improving the casting stability of low-silicon aluminum-containing steel according to claim 2, characterized in that: The process of determining the adjustable slag replacement cycle is as follows: If the target periodic sequence is stable, calculate the average period of all target slag replacement periods within the target periodic sequence and mark it as the adjustable slag replacement period; If the target periodic sequence is unstable, cluster analysis is performed on the target periodic sequence to obtain the adjustable slag period.

4. The method for improving the casting stability of low-silicon aluminum-containing steel according to claim 1, characterized in that: Determine whether the initial slag replacement cycle needs adjustment. If so, the process for adjusting the initial slag replacement cycle is as follows: Compare the adjustable slag replacement cycle with the initial slag replacement cycle: If the adjustable slag replacement cycle is greater than or equal to the initial slag replacement cycle, it is determined that no adjustment to the initial slag replacement cycle is required. Conversely, it is determined that the initial slag replacement cycle needs to be adjusted; When the initial slag replacement cycle needs to be adjusted, the initial slag replacement cycle will be updated and adjusted to be consistent with the adjustable slag replacement cycle.

5. The method for improving the casting stability of low-silicon aluminum-containing steel according to claim 4, characterized in that: The process of adjusting the volatility of the initial slag replacement cycle is as follows: During the historical slag replacement process, obtain the adjustable slag replacement cycle each time the initial slag replacement cycle is updated and adjusted. The cycle durations corresponding to the adjustable slag replacement cycle are integrated into a time-series cycle duration set according to the time sequence corresponding to the adjustable slag replacement cycle. Calculate the coefficient of variation of the time series period duration set to obtain the adjusted period fluctuation value; The adjustment cycle fluctuation value is compared with the adjustment cycle fluctuation threshold. If the adjustment cycle fluctuation value is greater than or equal to the adjustment cycle fluctuation threshold, it is determined that the adjustment volatility of the initial slag replacement cycle is large. Conversely, if the initial slag replacement cycle shows little fluctuation, it is determined that the adjustment volatility is small.

6. The method for improving the casting stability of low-silicon aluminum-containing steel according to claim 5, characterized in that: To determine whether there is a concentration of adjustments during the initial slag replacement cycle, and if so, to identify the objects of concentration, the process is as follows: Obtain the adjustable slag replacement cycle based on the initial slag replacement cycle when making cycle adjustments, and obtain the target cycle sequence corresponding to the adjustable slag replacement cycle. Summarize all target cycle sequences according to time sequence to obtain a mixed cycle sequence. Obtain the number of the first slag replacement cycle and the second slag replacement cycle in the mixed cycle sequence, and denot them as the first adjustment cycle number and the second adjustment cycle number; The first and second adjustment period numbers are respectively compared with the total number of periods of the mixed periodic sequence to obtain the first set of discriminant values ​​and the second set of discriminant values. The discriminant values ​​in the first and second ensembles are compared with the ensemble thresholds, respectively. If the first set of discrimination values ​​is greater than or equal to the set discrimination threshold, it indicates that there is an adjustment set phenomenon in the initial slag replacement cycle, and the set object is the first slag replacement cycle. If the second set of discrimination values ​​is greater than or equal to the set of discrimination threshold, it indicates that there is an adjustment set of values ​​in the initial slag replacement cycle, and the set of values ​​is the second slag replacement cycle. All other comparison cases were determined to be free of adjustment concentration phenomena in the initial slag replacement cycle.

7. The method for improving the casting stability of low-silicon aluminum-containing steel according to claim 6, characterized in that: The process of optimizing the adjustment method of the initial slag replacement cycle based on the adjustment focus is as follows: If the focus is on the first slag replacement cycle, then in the subsequent process of determining the adjustable slag replacement cycle, only the first slag replacement time and the working time of the protective slag are summed, and the summation result is used as the target slag replacement cycle. Through stability analysis, the adjustable slag replacement cycle is determined, and the initial slag replacement cycle is updated and adjusted to the adjustable slag replacement cycle. If the focus is on the second slag replacement cycle, then in the subsequent process of determining the adjustable slag replacement cycle, only the second slag replacement time and the working time of the protective slag are summed, and the summation result is used as the target slag replacement cycle. Through stability analysis, the adjustable slag replacement cycle is determined, and the initial slag replacement cycle is updated and adjusted to the adjustable slag replacement cycle.

Images