Intelligent detection method for liquid level height of side-blown converter based on machine learning

Abstract

The invention provides a side-blown converter liquid level height intelligent detection method based on machine learning, and the method comprises the steps: obtaining a pressure signal set in real time through a gas phase region pressure sensor, a slag phase pressure sensor and a copper matte phase pressure sensor which are disposed on the side wall of a side-blown converter; and importing the processed signal into a target liquid level height detection model based on machine learning training to generate liquid level height detection data. According to the model, the total height is calculated through a dynamic compensation formula by using three groups of pressure values, the formula comprises slag layer compensation, copper matte layer compensation and gas-liquid disturbance correction, and a temperature compensation coefficient and a turbulence correction coefficient are introduced. And a visual report is generated after detection data are confirmed, a redundant communication channel is enabled to recollect data when abnormity occurs, and accurate monitoring of the molten pool liquid level height is realized. According to the method, the precision and stability of intelligent detection of the liquid level height of the side-blown converter based on machine learning can be improved, and reliable data support is provided for optimization of the smelting process.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of automatic detection in the metallurgical industry, and more particularly, to a side-blown furnace liquid level intelligent detection method based on machine learning. BACKGROUND

[0002] In modern industrial production, the side-blown furnace, as an important metallurgical equipment, is widely used in the smelting process of non-ferrous metals. The liquid level height of the molten pool of the side-blown furnace is one of the key parameters affecting the smelting efficiency, product quality, and safe operation of the equipment. Accurate detection of the liquid level height of the molten pool is of great significance for optimizing the smelting process, improving production efficiency, reducing energy consumption, and ensuring stable operation of the equipment.

[0003] Currently, the detection methods for the liquid level height of the molten pool of the side-blown furnace mainly include manual measurement and automatic detection based on sensors. Manual measurement usually relies on the experience and intuitive judgment of the operator, which is not only inefficient and inaccurate, but also difficult to obtain real-time liquid level height information, and cannot meet the requirements of modern industrial production for precise control. Although the automatic detection technology based on sensors improves the detection efficiency and accuracy to some extent, the existing technology still has many limitations. For example, traditional sensor detection methods can only measure the pressure or liquid level height of a single phase region, and cannot fully reflect the pressure changes and liquid level height relationship in the gas phase region, slag phase, and copper matte phase of the molten pool. In addition, the existing technology also has deficiencies in data processing, such as inaccurate removal of interference data from the sensor signal, which affects the detection accuracy and makes it difficult to adapt to complex working conditions.

[0004] In the implementation of the embodiments of the present application, there are at least the following problems or defects in the prior art: the existing detection methods cannot comprehensively and accurately monitor the pressure changes and liquid level height relationship in different phase regions of the molten pool, the data processing accuracy is insufficient, it is difficult to adapt to complex working conditions, and it cannot generate real-time liquid level height detection data and visual reports, which cannot meet the requirements of modern industrial production for precise control and efficient management. SUMMARY

[0005] The present application provides a side-blown furnace liquid level intelligent detection method based on machine learning, comprising:

[0006] In response to determining that the detection processing time is reached, a set of pressure signals is obtained from pressure sensors deployed on the side wall of the side-blown furnace, the pressure sensors including a gas phase region pressure sensor located in the gas phase region, a slag phase pressure sensor located in the slag phase, and a copper matte phase pressure sensor located in the copper matte phase, to monitor the pressure changes in the gas phase region, slag phase, and copper matte phase regions of the furnace in real time;

[0007] The set of pressure signals is pre-processed to generate a set of pre-processed pressure signals;

[0008] The pre-processed pressure signal set is automatically imported into the associated file corresponding to the target liquid level detection model stored in the storage system to generate intelligent detection data of the liquid level of the side-blown furnace based on machine learning; wherein the target liquid level detection model is trained by a machine learning algorithm on a sample data set containing pressure data of the gas phase zone pressure sensor, the slag phase pressure sensor and the copper matte phase pressure sensor and corresponding actual liquid level of the molten pool;

[0009] According to the liquid level detection data, the height of the liquid level of the molten pool of the side-blown furnace is determined.

[0010] Further, the data preprocessing of the pressure signal set generates a pre-processed pressure signal set, which includes:

[0011] The interference data elimination operation is performed on the pressure signal set to obtain a pre-processed pressure signal set; wherein the interference data elimination operation includes:

[0012] The high-frequency noise caused by the strong molten pool agitation inherent in the side-blown process is eliminated from the slag phase pressure sensor signal;

[0013] The low-frequency drift caused by the fuel preheating system is corrected in the copper matte phase pressure sensor signal.

[0014] Further, the pre-processed pressure signal set is automatically imported into the associated file corresponding to the target liquid level detection model stored in the storage system to generate intelligent detection data of the liquid level of the side-blown furnace based on machine learning, which includes:

[0015] The pre-processed pressure signal set is automatically imported into the associated file corresponding to the target liquid level detection model to generate intelligent detection data of the liquid level of the side-blown furnace based on machine learning by using the data processing script contained in the target liquid level detection model.

[0016] Further, according to the liquid level detection data, the height of the liquid level of the molten pool of the side-blown furnace is determined, which includes:

[0017] The first pressure value P s3 measured by the gas phase zone pressure sensor in the pre-processed pressure signal set, the second pressure value P s2 measured by the slag phase pressure sensor, and the third pressure value P s1 measured by the copper matte phase pressure sensor are obtained.

[0018] The liquid level height H of the molten pool is calculated according to the following formula:

[0019]

[0020] Wherein: H is the liquid level height of the molten pool; hs is a dynamic slag layer height compensation value; h m is a dynamic copper matte layer height compensation value; Δh is a gas-liquid interaction disturbance correction amount; ρ s is a slag phase density; ρ m is a copper matte phase density, g is the acceleration of gravity; D is the inner diameter of the side-blown furnace; α(T) is a temperature compensation coefficient, which is a function of the bath temperature T, and is used to correct the deviation of the slag layer height calculation caused by temperature changes; β(T) is a temperature compensation coefficient, which is a function of the bath temperature T, and is used to correct the deviation of the copper matte layer height calculation caused by temperature changes; k1 is a turbulence correction coefficient, which is obtained by training a machine learning model, and is used to quantify the influence of the square of the pressure difference on the gas-liquid interaction disturbance; k2 is a turbulence correction coefficient, which is obtained by training a machine learning model, and is used to quantify the influence of the product of the two pressure differences on the gas-liquid interaction disturbance; P s3 is a first pressure value measured by a gas phase zone pressure sensor; P s2 is a second pressure value measured by a slag phase pressure sensor; P s1 is a third pressure value measured by a copper matte phase pressure sensor.

[0021] Further, the automatic import of the pre-processed pressure signal set into the associated file corresponding to the target liquid level height detection model stored in the storage system generates side-blown furnace liquid level height intelligent detection data based on machine learning, which includes:

[0022] In response to the total number of pressure sensors being lower than a first value, the pre-processed pressure signal set is imported into the associated file of the storage system to generate liquid level height detection data;

[0023] In response to the total number of pressure sensors being higher than the first value, the following operations are performed for each detection area of the side-blown furnace: in response to the number of sensors in the current detection area being higher than a second value, the pressure signal set of the current area is grouped to obtain a pressure signal group set, the number of sensors in each group being less than the second value, and the second value being less than a division value, the division value being the first value divided by the number of detection areas;

[0024] In response to the number of sensors in the current detection area being lower than or equal to the second value, the pressure signal set of the current area is determined as the pressure signal group set;

[0025] Determine the total number of groups of all pressure signal groups;

[0026] Assign a unique group identifier to each pressure signal group;

[0027] Create an associated file in the storage system, the number of which is equal to the total number of groups;

[0028] Set the file name of all associated files to the corresponding group identifier;

[0029] creating a number of copies of the target liquid level detection model equal to the total number of groups in the storage system;

[0030] binding each copy of the model to an associated file with a corresponding file name;

[0031] dividing the pre-processed pressure signal set into a set of pressure signal data groups and a set of group identification data groups according to the grouping result;

[0032] generating liquid level detection data based on the set of pressure signal data groups, the set of group identification data groups, the copies of the model, and the associated files.

[0033] Further, the generating liquid level detection data based on the set of pressure signal data groups, the set of group identification data groups, the copies of the model, and the associated files comprises:

[0034] for each pressure signal group:

[0035] importing the pressure signal data and group identification data of the group into the associated file with the matching file name;

[0036] calling the bound copy of the model to generate a pressure signal detection data group;

[0037] collecting all pressure signal detection data groups and generating liquid level detection data using the area analysis script included in the target liquid level detection model.

[0038] Further, the target liquid level detection model supports using gradient boosting decision tree algorithm for molten pool liquid level prediction.

[0039] Further, it further comprises:

[0040] obtaining liquid level detection confirmation information returned by the side-blown furnace operation terminal;

[0041] in response to confirming that the detection data is correct, generating a visualization report containing the side-blown furnace identifier, liquid level detection data, and liquid level information;

[0042] sending the visualization report to the monitoring terminal.

[0043] Further, the generating a visualization report comprises:

[0044] for each detection area of the gas phase, slag phase, and copper matte phase:

[0045] extracting the liquid level detection data of the current area;

[0046] generating local liquid level and area liquid value information according to the current area data, the area liquid value information being calculated by the following formula:

[0047] Vs = η · (h s -h sopt ) 2 · Q cu

[0048] V m = μ · (h m -h mopt ) 2 · Q cu

[0049] Wherein: V s is the metallurgical value loss of the slag phase zone; V m is the metallurgical value loss of the copper matte zone; η is the slag phase zone metallurgical efficiency coefficient, used to measure the influence degree of the liquid level of the slag phase zone deviating from the optimal value on the metallurgical value; μ is the copper matte zone metallurgical efficiency coefficient, used to measure the influence degree of the liquid level of the copper matte zone deviating from the optimal value on the metallurgical value; h sopt is the optimal phase layer height of the slag phase zone; h mopt is the optimal phase layer height of the copper matte zone; Q cu is the copper matte yield; h s is the dynamic slag layer height compensation value; h m is the dynamic copper matte layer height compensation value.

[0050] Further, the automatic importing of the pretreatment pressure signal set to the associated file corresponding to the target liquid level detection model stored in the storage system generates intelligent detection data of the liquid level of the side-blown furnace based on machine learning, and further comprises:

[0051] In response to the transmission of the pressure sensor being abnormal, a redundant communication channel is enabled to reacquire the pressure signal set, and the process of acquiring the pressure signal to determining the liquid level is repeatedly executed.

[0052] The above embodiments of the present application have at least the following beneficial effects:

[0053] 1. Through the cooperative monitoring of the three groups of pressure sensors of the gas phase zone, the slag phase and the copper matte, combined with the dynamic compensation algorithm and the temperature correction coefficient, the measurement error problem caused by the stratified flow and temperature fluctuation of the molten pool in the traditional single sensor detection is effectively solved, and the precision and reliability of the liquid level detection are improved.

[0054] 2. The target liquid level detection model trained by machine learning is adopted, combined with the interference data elimination and the redundant communication mechanism, the interference of the high-frequency noise of the oxygen-enriched side blowing, the low-frequency drift of the fuel preheating and the abnormal sensor signal on the detection process is overcome, and the anti-interference ability and fault tolerance of the system are enhanced.

[0055] 3. Through dynamic calculation of the slag layer height compensation value, the copper matte layer height compensation value and the gas-liquid interaction disturbance correction amount, and generation of a visual report containing metallurgical value loss analysis, real-time optimization monitoring of the molten pool liquid level is realized, accurate data support is provided for smelting process adjustment, and production efficiency and resource utilization are improved. BRIEF DESCRIPTION OF DRAWINGS

[0056] The above and other objects, features and advantages of the present application will become more apparent from the following detailed description read in conjunction with the accompanying drawings. In the drawings, several embodiments of the present application are shown by way of example, and not limitation, wherein:

[0057] Figure 1 A flowchart of a machine learning-based intelligent detection method for liquid level height of a side-blown furnace according to an embodiment of the present application is provided.

[0058] Figure 2 A schematic diagram of a side-blown furnace pool according to an embodiment of the present application is provided.

[0059] Figure 3 A schematic diagram of the structure of an electronic device according to an embodiment of the present application is shown. DETAILED DESCRIPTION

[0060] The principles and spirits of the present application will be described below with reference to several exemplary embodiments. It should be understood that these embodiments are given only to enable those skilled in the art to better understand and implement the present application, and do not limit the scope of the present application in any way. On the contrary, these embodiments are provided to make the present application more thorough and complete, and to fully convey the scope of the present application to those skilled in the art.

[0061] Those skilled in the art know that the embodiments of the present application can be implemented as a system, device, apparatus, method or computer program product. Therefore, the present application can be embodied as a complete hardware, complete software (including firmware, resident software, microcode, etc.), or a combination of hardware and software.

[0062] It should be noted that any number of elements in the drawings is used for example and not limitation, and any naming is only for distinction and does not have any limiting meaning.

[0063] Reference will now be made to Figure 1 , Figure 1 A flowchart of a machine learning-based intelligent detection method for liquid level height of a side-blown furnace according to an embodiment of the present application is provided. As shown in Figure 1 , a machine learning-based intelligent detection method for liquid level height of a side-blown furnace includes:

[0064] Step 1: In response to determining that the detection processing time has been reached, a pressure signal set is acquired from the pressure sensors deployed on the side wall of the side-blown furnace. The pressure sensors include a gas phase pressure sensor located in the gas phase zone, a slag phase pressure sensor located in the slag phase, and a copper matte phase pressure sensor located in the copper matte phase, so as to monitor the pressure changes in the gas phase zone, slag phase, and copper matte phase regions in the furnace in real time.

[0065] Step 2: Perform data preprocessing on the pressure signal set to generate a preprocessed pressure signal set;

[0066] Step 3: Automatically import the preprocessed pressure signal set into the associated file stored in the storage system corresponding to the target liquid level height detection model to generate intelligent detection data of side-blown furnace liquid level height based on machine learning; wherein, the target liquid level height detection model is trained on the sample dataset by machine learning algorithm, and the sample dataset includes pressure data from the gas phase pressure sensor, slag phase pressure sensor and copper matte phase pressure sensor and the corresponding actual liquid level height of the molten pool;

[0067] Step 4: Determine the height of the molten pool in the side-blown furnace based on the liquid level detection data.

[0068] It should be noted that this invention proposes an intelligent method for detecting the molten metal level in a side-blown furnace based on machine learning. The core of this method lies in acquiring pressure signal sets through pressure sensors deployed on the sidewall of the side-blown furnace, and then using a target molten metal level detection model trained by a machine learning algorithm to accurately detect the molten metal level. A side-blown furnace is a device used for smelting non-ferrous metals. It contains three regions: a gas phase, a slag phase, and a copper matte phase. Pressure changes in each region are closely related to the molten metal level. Pressure sensors are devices used to monitor pressure changes within the furnace in real time. By acquiring these pressure signals, the height changes of the molten metal level can be indirectly reflected. Data preprocessing refers to the preliminary processing of the acquired pressure signals to remove noise and interference data, thereby improving detection accuracy. The target molten metal level detection model is trained using a machine learning algorithm on a sample dataset. The sample dataset contains pressure data from different pressure sensors and the corresponding actual molten metal level. The model can predict the molten metal level based on the input pressure signal.

[0069] Specifically, the pressure sensors include a gas phase pressure sensor located in the gas phase, a slag phase pressure sensor located in the slag phase, and a copper matte phase pressure sensor located in the copper matte phase. The gas phase pressure sensor monitors pressure changes in the gas phase, the slag phase pressure sensor monitors pressure changes in the slag phase, and the copper matte phase pressure sensor monitors pressure changes in the copper matte phase. The pressure data from these sensors correspond to the pressure signals in the gas phase, slag phase, and copper matte phase, respectively. Data preprocessing includes removing interfering data, such as high-frequency noise from the slag phase pressure sensor signal, which may be caused by the oxygen-enriched side-blowing system; and correcting low-frequency drift in the copper matte phase pressure sensor signal, which may be caused by the fuel preheating system. The preprocessed pressure signal set is then imported into a target liquid level detection model, which is trained using a machine learning algorithm and can calculate the molten pool level height based on the input pressure signal. The model's input parameters include the first pressure value P3 measured by the gas phase pressure sensor, the second pressure value P2 measured by the slag phase pressure sensor, and the third pressure value P1 measured by the copper matte phase pressure sensor. These pressure values ​​have a certain mathematical relationship with the height of the molten pool.

[0070] Preferably, the construction process of the target liquid level height detection model is as follows: First, a large number of sample datasets are collected. These datasets include pressure data from gas phase pressure sensors, slag phase pressure sensors, and copper matte phase pressure sensors, as well as the corresponding actual liquid level height of the molten pool. Then, machine learning algorithms are used to train these sample data. The algorithm learns the relationship between pressure data and liquid level height to generate a model capable of predicting liquid level height. In the data preprocessing stage, filtering algorithms can be used to remove interfering data. For example, a low-pass filter is used to remove high-frequency noise, and a high-pass filter is used to correct low-frequency drift. When calculating the liquid level height, the model calculates the liquid level height using a specific mathematical formula based on the pressure values ​​in the preprocessed pressure signal set, combined with the physical parameters of the molten pool, such as slag phase density, copper matte phase density, gravitational acceleration, side-blown furnace inner diameter, and temperature compensation coefficient. The selection of these parameters and coefficients is based on the physical characteristics of the molten pool and actual working conditions, obtained through experiments and data analysis to ensure the accuracy and reliability of the calculation results.

[0071] In some embodiments, the step of preprocessing the pressure signal set to generate a preprocessed pressure signal set includes:

[0072] An interference data removal operation is performed on the pressure signal set to obtain a preprocessed pressure signal set; wherein, the interference data removal operation includes:

[0073] Eliminate the high-frequency noise in the slag phase pressure sensor signal caused by the strong molten pool agitation inherent in the side-blowing process;

[0074] Correct the low-frequency drift in the copper matte phase pressure sensor signal caused by the fuel preheating system.

[0075] It should be noted that data preprocessing of the pressure signal set is one of the key steps in ensuring the accuracy of liquid level detection in this invention. The purpose of data preprocessing is to remove noise and abnormal data that may interfere with the detection results, thereby improving the accuracy and reliability of the detection. Specifically, the interference data removal operation includes removing high-frequency noise from the slag phase pressure sensor signal and correcting low-frequency drift in the copper matte phase pressure sensor signal. High-frequency noise is usually caused by rapid pressure fluctuations generated during the operation of the oxygen-enriched side-blowing system, while low-frequency drift may be caused by slow changes in pressure sensor readings due to the fuel preheating system. Through these preprocessing operations, the pressure signal can be made more stable and accurate, providing a reliable data foundation for subsequent liquid level calculation.

[0076] Specifically, interference data removal in data preprocessing refers to the process of filtering and correcting the acquired pressure signals. High-frequency noise in slag-phase pressure sensor signals refers to rapidly changing pressure fluctuations over a short period, which may mask the true liquid level information. This noise is typically caused by airflow disturbances generated during the operation of the oxygen-enriched side-blowing system. Low-frequency drift in copper-matte-phase pressure sensor signals refers to the phenomenon of slowly changing pressure readings over time, which may be due to zero-point drift of the pressure sensor caused by the fuel preheating system. To remove these interfering data, signal processing algorithms can be used, such as low-pass filters to remove high-frequency noise and high-pass filters to correct low-frequency drift. The parameter settings of these algorithms need to be determined based on the actual operating conditions and sensor characteristics; for example, the filter cutoff frequency can be selected based on the frequency range of the noise and the characteristic frequency of the signal.

[0077] Preferably, the specific implementation steps for interference data removal are as follows: First, analyze the slag phase pressure sensor signal to determine the frequency range of high-frequency noise. Then, select appropriate low-pass filter parameters; for example, the cutoff frequency can be set below the upper limit of the noise frequency range to ensure that high-frequency noise is effectively filtered out while retaining useful information in the signal. For low-frequency drift in the copper matte phase pressure sensor signal, it can be corrected using a high-pass filter. The cutoff frequency of the filter can be set according to the drift frequency range and the characteristic frequency of the signal. Furthermore, time series analysis methods can be combined to perform trend analysis on the pressure signal to further correct low-frequency drift. Through these processing steps, the pre-processed pressure signal can be ensured to be more stable and accurate, providing reliable data support for subsequent liquid level detection.

[0078] In some embodiments, the step of automatically importing the preprocessed pressure signal set into the associated file corresponding to the target liquid level detection model and stored in the storage system to generate intelligent detection data for the liquid level of a side-blown furnace based on machine learning includes:

[0079] The preprocessed pressure signal set is automatically imported into the associated file corresponding to the target liquid level detection model, so as to generate intelligent detection data of side-blown furnace liquid level based on machine learning by utilizing the data processing script contained in the target liquid level detection model.

[0080] It should be noted that the automatic importation of the preprocessed pressure signal set into the associated file corresponding to the target liquid level detection model, as mentioned in this invention, is one of the key steps in achieving automated liquid level detection. This process utilizes the data processing script contained in the target liquid level detection model, which can efficiently generate intelligent detection data for the liquid level of a side-blown furnace based on machine learning. The target liquid level detection model is trained based on a machine learning algorithm, and its purpose is to accurately predict the molten pool level by processing pressure signals. The associated file refers to the data file stored in the storage system that is associated with the target liquid level detection model. It is used to store the preprocessed pressure signal set so that the model can read and process this data.

[0081] Specifically, the preprocessed pressure signal set refers to the collection of pressure signals after data preprocessing. These signals have had interference data removed and can more accurately reflect the pressure changes inside the side-blown furnace. The target liquid level height detection model is a trained machine learning model that can calculate the height of the molten pool based on the input pressure signals. The data processing script is the program code in the model used to process the input data; it defines how to convert the preprocessed pressure signal set into liquid level height detection data. The storage system refers to the hardware or software system used to store the data and model; it can be a local server, cloud storage, or other forms of data storage facilities. The associated file is the file in the storage system corresponding to the target liquid level height detection model, used to store the preprocessed pressure signal set so that the model can read and process this data.

[0082] Preferably, the construction process of the target liquid level height detection model is as follows: First, a large number of sample datasets are collected, which include pressure sensor data for the gas phase, slag phase, and copper matte phase, as well as the corresponding actual liquid level height of the molten pool. Then, machine learning algorithms are used to train these sample datasets. The algorithm learns the relationship between pressure data and liquid level height to generate a model capable of predicting liquid level height. In the data processing script, the input parameters include a preprocessed set of pressure signals. After processing by the script, these signals generate liquid level height detection data. The script processing may include steps such as data normalization and feature extraction to ensure that the data format input to the model is correct and representative. For example, data normalization can adjust the numerical range of pressure signals to a uniform standard so that the model can process these data more effectively. Feature extraction can extract key features related to liquid level height from the original pressure signals, such as pressure difference and pressure change rate. Through these processing steps, the model can calculate the liquid level height more accurately, thereby providing reliable data support for the operation of the side-blown furnace.

[0083] In some embodiments, determining the height of the molten pool in a side-blown furnace based on the liquid level detection data includes:

[0084] The first pressure value P measured by the gas phase region pressure sensor in the preprocessed pressure signal set is obtained. s3 The second pressure value P measured by the slag phase pressure sensor s2 The third pressure value P measured by the copper matte phase pressure sensor s1 ;

[0085] Calculate the molten pool level H using the following formula:

[0086]

[0087] Where H is the height of the molten pool; h s h is the dynamic slag layer height compensation value. m ρ is the dynamic copper matte layer height compensation value; Δh is the gas-liquid interaction disturbance correction amount; s ρ is the density of the slag phase; m ρ is the density of the copper matte phase; g is the acceleration due to gravity; D is the inner diameter of the side-blown furnace; α(T) is the temperature compensation coefficient, a function of the molten pool temperature T, used to correct the deviation in slag layer height calculation caused by temperature changes; β(T) is the temperature compensation coefficient, a function of the molten pool temperature T, used to correct the deviation in copper matte layer height calculation caused by temperature changes; k1 is the turbulence correction coefficient, obtained through machine learning model training, used to quantify the influence of the square of the pressure difference on gas-liquid interaction disturbance; k2 is the turbulence correction coefficient, obtained through machine learning model training, used to quantify the influence of the product of the two pressure differences on gas-liquid interaction disturbance; P s3The first pressure value measured by the gas phase pressure sensor; P s2 The second pressure value measured by the slag phase pressure sensor; P s1 This is the third pressure value measured by the copper matte phase pressure sensor.

[0088] It should be noted that the process of determining the molten pool height in the side-blown furnace based on liquid level detection data, as mentioned in this invention, is the core step of the entire detection method. This process acquires the pressure values ​​from the pre-processed pressure signal set and calculates the molten pool height using a specific formula, thereby achieving accurate measurement of the molten pool height in the side-blown furnace. This formula comprehensively considers the pressure difference between the gas phase, slag phase, and copper matte phase, as well as relevant physical parameters and correction coefficients, to ensure the accuracy and reliability of the calculation results. The settings of these parameters and coefficients are based on the physical characteristics and actual operating conditions of the side-blown furnace, and can effectively compensate for the influence of factors such as temperature changes and gas-liquid interaction disturbances on the liquid level height calculation.

[0089] Specifically, the first pressure value in the pre-processed pressure signal set refers to the pressure value measured by the gas phase pressure sensor, reflecting the pressure state of the gas phase; the second pressure value refers to the pressure value measured by the slag phase pressure sensor, reflecting the pressure state of the slag phase; and the third pressure value refers to the pressure value measured by the copper matte phase pressure sensor, reflecting the pressure state of the copper matte phase. These pressure values ​​are the basic data for calculating the molten pool height. The dynamic slag layer height compensation value and the dynamic copper matte layer height compensation value in the formula are parameters dynamically adjusted according to actual working conditions, used to compensate for changes in the height of the slag layer and copper matte layer. The gas-liquid interaction disturbance correction amount is used to correct the interaction influence between the gas and liquid interfaces. The physical parameters involved in the formula include slag phase density, copper matte phase density, gravitational acceleration, and the inner diameter of the side-blown furnace, etc. The values ​​of these parameters can be obtained through experimental measurement or by consulting relevant materials. The temperature compensation coefficient is a function of the molten pool temperature, used to correct the calculation deviation of the slag layer and copper matte layer height caused by temperature changes. The turbulence correction coefficient is obtained by training a machine learning model and is used to quantify the impact of pressure difference on gas-liquid interaction disturbances.

[0090] Preferably, the calculation process for the liquid level height can be further refined as follows: First, extract the pressure values ​​of the gas phase, slag phase, and copper matte phase from the pre-processed pressure signal set. Then, calculate the molten pool liquid level height according to the formula. The dynamic compensation value in the formula can be dynamically adjusted according to the real-time monitored state of the slag layer and copper matte layer, for example, based on changes in slag layer thickness or the flow of the copper matte layer. The temperature compensation coefficient can be obtained by fitting experimental data, for example, by measuring the height changes of the slag layer and copper matte layer at different temperatures, and then establishing a functional relationship between temperature and the compensation coefficient. The training process for the turbulence correction coefficient may include: collecting a large amount of pressure data and corresponding liquid level height data under different operating conditions, using machine learning algorithms, such as linear regression and neural networks, to train these data, thereby obtaining correction coefficients that can quantify the influence of pressure difference. In actual calculations, the values ​​of these parameters and coefficients are substituted into the formula, and the molten pool liquid level height is finally obtained by calculating the pressure difference and its combination relationship. This process not only considers the changes in pressure signals, but also improves the accuracy and adaptability of the liquid level height calculation through compensation and correction mechanisms.

[0091] In some embodiments, the step of automatically importing the preprocessed pressure signal set into the associated file corresponding to the target liquid level detection model and stored in the storage system to generate intelligent detection data for the liquid level of a side-blown furnace based on machine learning includes:

[0092] In response to the total number of pressure sensors being less than a first value, the preprocessed pressure signal set is imported into the associated file of the storage system to generate liquid level detection data;

[0093] In response to the total number of pressure sensors being higher than the first value, the following operations are performed for each detection area of ​​the side-blown furnace: In response to the number of sensors in the current detection area being higher than the second value, the pressure signal set of the current area is grouped to obtain a pressure signal set, the number of sensors in each group is less than the second value, and the second value is less than the division value, the division value being the first value divided by the number of detection areas.

[0094] In response to the number of sensors in the current detection area being less than or equal to the second value, the pressure signal set in the current area is determined as a pressure signal set;

[0095] Determine the total number of pressure signal sets;

[0096] Assign a unique group identifier to each pressure signal group;

[0097] Create an association file in the storage system with a number equal to the total number of groups;

[0098] Set the filenames of all associated files to the corresponding group identifier;

[0099] Create a number of copies of the target liquid level height detection model in the storage system equal to the total number of groups;

[0100] Bind each model copy to an associated file with the corresponding filename;

[0101] Based on the grouping results, the preprocessed pressure signal set is divided into a pressure signal data set and a group identifier data set.

[0102] Based on the pressure signal data set, group identifier data set, model copy, and associated files, liquid level detection data is generated.

[0103] It should be noted that the process mentioned in this invention, which automatically imports the preprocessed pressure signal set into the associated file of the storage system corresponding to the target liquid level detection model and generates liquid level detection data, is to improve detection efficiency and flexibility to adapt to different sensor configurations. When the number of pressure sensors changes, the system can automatically adjust the data processing method to ensure the accuracy and reliability of liquid level detection. This process specifically considers the cases of insufficient or excessive number of sensors, and optimizes the data processing flow through grouping and model replication mechanisms.

[0104] Specifically, the total number of pressure sensors refers to the total number of pressure sensors deployed on the sidewall of the side-blown furnace, including those for the gas phase, slag phase, and copper matte phase. When the number of sensors is below a certain first set value, the system directly imports the pre-processed pressure signals into the associated file in the storage system to generate liquid level detection data. When the number of sensors exceeds the first value, the system groups the signals according to the number of sensors in each detection area. If the number of sensors in a detection area exceeds a second value, the system groups the pressure signals for that area, with each group containing fewer sensors than the second value, which is less than a division value obtained by dividing the first value by the number of detection areas. In this way, the system can flexibly handle different numbers of sensor signals, ensuring efficient and accurate data processing.

[0105] Preferably, when the number of sensors is large, the system groups each detection area. The specific steps are as follows: First, determine if the number of sensors in each detection area is higher than a second value. If it is, group the pressure signals for that area, with each group containing fewer than the second value. If the number is lower than or equal to the second value, treat the pressure signal set for that area as a single group. Then, the system determines the total number of groups and assigns a unique group identifier to each group. Next, create an equal number of associated files in the storage system as the total number of groups, and name these files with the corresponding group identifiers. Simultaneously, create an equal number of target liquid level detection model copies in the storage system as the total number of groups, and bind each model copy to its corresponding associated file. Finally, based on the grouping results, divide the preprocessed pressure signal set into multiple pressure signal data groups and group identifier data groups. Then, based on these data groups, model copies, and associated files, generate liquid level detection data. Through this grouping and model copying mechanism, the system can efficiently process a large number of sensor signals, ensuring the accuracy and flexibility of liquid level detection.

[0106] In some embodiments, generating liquid level detection data based on the pressure signal data set, group identifier data set, model copy, and associated files includes:

[0107] Perform the following for each pressure signal group:

[0108] Import the pressure signal data and group identification data of this group into the associated file with the matching filename;

[0109] The bound model copy is invoked to generate a pressure signal detection data set;

[0110] By aggregating all pressure signal detection data sets and utilizing the region analysis script contained in the target liquid level detection model, liquid level detection data is generated.

[0111] It should be noted that the process mentioned in this invention for generating liquid level detection data based on pressure signal data sets, group identifier data sets, model copies, and associated files is an optimized processing method for intelligent liquid level detection systems in side-blown furnaces with a large number of sensors. This process, by grouping pressure signals and assigning an independent model copy and associated file to each group, can efficiently process large amounts of sensor data while ensuring the accuracy and real-time performance of liquid level detection. This method is particularly suitable for complex operating conditions where sensors are unevenly distributed or numerous, effectively improving the system's flexibility and reliability.

[0112] Specifically, the pressure signal data set refers to the data collection formed by dividing the preprocessed pressure signals according to the grouping results. Each group corresponds to a portion of the sensor data in a detection area. The group identifier data set is a set of unique identifiers assigned to each pressure signal group to distinguish different groups. The model copy is multiple instances of the target liquid level detection model. Each copy is bound to an associated file and is used to process the pressure signal data of its corresponding group. The associated file is a file stored in the storage system to store the pressure signal data of each group and the corresponding detection results. During the generation of liquid level detection data, the system imports the pressure signal data and group identifier data of each group into the corresponding associated file, and then calls the bound model copy for processing, ultimately generating the liquid level detection data. This method, through parallel computation of group processing and model copies, can effectively improve data processing efficiency and adapt to the processing needs of large-scale sensor data.

[0113] Preferably, the specific steps for generating liquid level detection data can be further refined as follows: First, for each pressure signal group, the pressure signal data and corresponding group identifier data of that group are imported into an associated file with a matching filename. Then, the model copy bound to the associated file is called to process the imported data and generate the pressure signal detection data for that group. The processing of the model copy may include steps such as data normalization, feature extraction, and model prediction. The specific implementation of these steps can be determined according to the model design and training process. For example, data normalization can adjust the numerical range of the pressure signals to a unified standard so that the model can process these data more effectively; feature extraction can extract key features related to liquid level, such as pressure difference and pressure change rate, from the original pressure signals. Finally, the pressure signal detection data of all groups are collected, and the data are comprehensively analyzed using the region analysis script contained in the target liquid level detection model to generate the final liquid level detection data. The role of the region analysis script is to integrate the detection results of different groups, consider the relationship between different detection areas, and thus obtain the liquid level distribution of the entire side-blown furnace molten pool. This method can make full use of the data of each group and improve the accuracy and reliability of liquid level detection.

[0114] In some embodiments, the target liquid level detection model supports the use of a gradient boosting decision tree algorithm for predicting the liquid level of the molten pool.

[0115] It should be noted that the target liquid level detection model mentioned in this invention supports the use of the gradient boosting decision tree algorithm for predicting the molten pool liquid level. This technical solution aims to improve the accuracy and reliability of liquid level detection through advanced machine learning algorithms. The gradient boosting decision tree algorithm is an ensemble learning method that can effectively handle complex nonlinear relationships and noisy data by constructing multiple decision tree models and progressively optimizing the prediction results. In the intelligent detection of liquid level in a side-blown furnace based on machine learning, this algorithm can fully utilize the feature information in the sensor data to accurately predict the liquid level, while also exhibiting strong robustness to outliers and noise in the data.

[0116] Specifically, the gradient boosting decision tree algorithm is an ensemble learning method based on gradient boosting. It improves the overall predictive performance of the model by progressively constructing multiple decision tree models and optimizing the gradient direction of the prediction error at each step. In this invention, the algorithm is used to process a preprocessed pressure signal set obtained from pressure sensors and predict the molten pool level. The model's input parameters include pressure sensor data from the gas phase, slag phase, and copper matte phase. After preprocessing, this data more accurately reflects the pressure changes within the molten pool. The model's output is the predicted molten pool level, obtained by comprehensively considering data from multiple pressure sensors and related physical parameters such as density and temperature. During model training, the sample dataset contains known molten pool levels and their corresponding pressure sensor data. The model trained using this data can accurately predict unknown molten pool levels in practical applications.

[0117] Preferably, the construction process of the gradient boosting decision tree model can be further refined as follows: First, a large number of sample datasets are collected, including pressure sensor data for the gas phase, slag phase, and copper matte phase, as well as the corresponding actual liquid level height of the molten pool. Then, these data are preprocessed, including data cleaning and normalization, to ensure data quality and consistency. Next, the gradient boosting decision tree algorithm is used to train the preprocessed data. During training, the algorithm gradually builds multiple decision trees, each optimized based on the previous one to reduce prediction errors. Specific training steps include: initializing the model's predicted values, calculating residuals, constructing new decision trees to fit the residuals, updating the model's predicted values, and repeating the above steps until a preset number of iterations is reached or the model performance no longer improves. After model training is complete, the model's performance can be evaluated using methods such as cross-validation, and model parameters, such as the learning rate, tree depth, and number of trees, can be adjusted as needed. The final model can accurately predict the molten pool level height based on the input pressure sensor data, providing reliable data support for the operation monitoring of the side-blown furnace.

[0118] In some embodiments, it also includes:

[0119] Obtain the liquid level detection confirmation information returned by the side-blown furnace operation terminal;

[0120] In response to confirmation that the test data is correct, a visual report is generated that includes the side-blown furnace identification, liquid level test data, and liquid level information;

[0121] The visualization report is sent to the monitoring terminal.

[0122] It should be noted that the acquisition of liquid level detection confirmation information returned by the side-blown furnace operating terminal, and the generation of a visual report containing the side-blown furnace identifier, liquid level detection data, and liquid level information, mentioned in this invention, is a crucial step in ensuring the accuracy of the detection results and facilitating monitoring and management. The side-blown furnace operating terminal refers to the equipment used to operate and monitor the side-blown furnace; operators can confirm the detection results through this terminal. The visual report is a way to intuitively display detection data and results, helping operators quickly understand the real-time status of the molten pool liquid level in the side-blown furnace, thereby allowing for timely adjustments to process parameters and ensuring the stability and safety of the production process.

[0123] Specifically, liquid level detection confirmation information refers to the feedback information returned by the operating terminal regarding the accuracy of the liquid level detection data. Operators can confirm the reliability of the detection system's data by comparing it with actual observation data or the results of other detection methods. If the data is confirmed to be correct, the system will generate a visual report. This report includes the side-blown furnace identifier, a unique identifier used to distinguish different side-blown furnaces, the liquid level detection data, the actual detected liquid level value, and liquid level information such as the trend of liquid level changes and alarm information. This information is displayed through graphics, tables, or other visualization methods, enabling operators to quickly understand the detection results. For example, liquid level information can be displayed as a line graph showing the trend of liquid level changes over time, or color-coded to indicate whether the liquid level is within the normal range.

[0124] Preferably, the process of generating a visualization report can be further refined as follows: First, the system obtains liquid level detection confirmation information from the side-blown furnace operation terminal. If the confirmation information indicates that the detection data is correct, the system will begin generating a visualization report. The report generation process includes: extracting identification information related to the current side-blown furnace from the database. This identification information can be the furnace number, furnace type, etc., used to identify the side-blown furnace corresponding to the report; extracting liquid level detection data, which are accurate values ​​calculated and verified by the model; generating liquid level information based on the liquid level data, for example, by analyzing the trend of liquid level changes to determine whether there are abnormal fluctuations. Next, the system integrates this information into a visualization interface, for example, using charts to display the change of liquid level over time, or using text to describe the current state and possible trends of the liquid level. Finally, the system sends the generated visualization report to the monitoring terminal, where operators can view the report content in real time and make timely decisions. This visualization report not only improves the readability of the data but also enhances the operator's control over the production process.

[0125] In some embodiments, generating a visualization report includes:

[0126] The following procedures were performed on each detection area: gas phase, slag phase, and copper matte phase.

[0127] Extract the liquid level detection data for the current area;

[0128] Based on the current regional data, local liquid level height and regional liquid level value information are generated. The regional liquid level value information is calculated using the following formula:

[0129] V s =η·(h s -h sopt ) 2 ·Q cu

[0130] V m =μ·(h m -h mopt ) 2 ·Q cu

[0131] Among them, V s For the metallurgical value loss in the slag phase zone; V m η is the metallurgical value loss in the copper matte zone; η is the metallurgical efficiency coefficient in the slag phase zone, used to measure the impact of the liquid level in the slag phase zone deviating from the optimal value on the metallurgical value; μ is the metallurgical efficiency coefficient in the copper matte zone, used to measure the impact of the liquid level in the copper matte zone deviating from the optimal value on the metallurgical value; h sopt The optimal phase layer height in the slag phase zone; h mopt The optimal phase height for the copper matte region; Qcu For copper matte production; h s h is the dynamic slag layer height compensation value. m This is the dynamic copper matte layer height compensation value.

[0132] It should be noted that the process of generating a visual report mentioned in this invention, particularly the processing of liquid level detection data for each detection area in the gas phase, slag phase, and copper matte phase, and the calculation of regional liquid level value information, is a crucial step in providing operators with more comprehensive and intuitive monitoring information. By extracting the liquid level detection data for each area and calculating the metallurgical value loss in the slag phase and copper matte phases, the impact of liquid level deviation from the optimal value on production efficiency and product quality can be assessed more accurately. This method not only helps operators monitor liquid level in real time but also provides a basis for decision-making to optimize the production process through the quantification of value loss.

[0133] Specifically, regional liquid level value information refers to the quantitative information on the metallurgical value loss caused by deviations in liquid level height from the optimal value in the slag phase and copper matte zones, obtained through calculation. The metallurgical value loss in the slag phase and copper matte zones is calculated using specific formulas that consider the degree of deviation from the optimal liquid level height and the corresponding metallurgical efficiency coefficient. The metallurgical efficiency coefficient is a parameter that measures the impact of liquid level height deviation on metallurgical value, reflecting the specific impact of liquid level height changes on production efficiency and product quality. The optimal phase layer height refers to the best values ​​of liquid level height in the slag phase and copper matte zones under ideal operating conditions. These values ​​are obtained experimentally or empirically and used as target values ​​for liquid level height control. Copper matte production is an important production parameter that directly affects the calculation of metallurgical value loss. By calculating these parameters, the metallurgical value loss in the slag phase and copper matte zones can be obtained, thus providing data support for optimizing the production process.

[0134] Preferably, the process of generating the visualization report can be further refined as follows: First, the system extracts liquid level detection data for each detection area in the gas phase, slag phase, and copper matte phase from the storage system. Then, based on this data, the metallurgical value loss in the slag phase and copper matte phase is calculated. The specific calculation process includes: determining the degree to which the liquid level deviates from the optimal value, i.e., the difference between the current liquid level and the optimal phase layer height; based on this difference, combined with the metallurgical efficiency coefficient and copper matte production, the metallurgical value loss is calculated using a specific formula. For example, if the liquid level deviates significantly from the optimal value, the metallurgical efficiency coefficient will reflect a larger value loss. After the calculation is completed, the system integrates this information into the visualization report and presents it to the operators in an intuitive way. For example, the metallurgical value loss in different areas can be displayed through a bar chart, or specific values ​​and corresponding liquid level heights can be listed in a table. In addition, the report can also include a real-time trend chart of the liquid level change and a comparison with the optimal phase layer height to help operators understand the current production status more intuitively. This detailed visualization report not only provides real-time data on liquid level, but also offers operators a basis for optimizing production decisions by quantifying value loss.

[0135] In some embodiments, the step of automatically importing the preprocessed pressure signal set into the associated file corresponding to the target liquid level detection model and stored in the storage system to generate intelligent detection data of side-blown furnace liquid level based on machine learning further includes:

[0136] In response to a pressure sensor transmission anomaly, a redundant communication channel is activated to reacquire the pressure signal set, and the process of acquiring the pressure signal and determining the liquid level is repeated.

[0137] It should be noted that the method described in this invention, which involves activating a redundant communication channel to reacquire the pressure signal set when the pressure sensor transmission fails, and repeating the process from pressure signal acquisition to liquid level determination, is to ensure the reliability and stability of the detection system. In industrial environments, sensor data transmission may be affected by factors such as electromagnetic interference, network failures, or equipment malfunctions, leading to data loss or errors. By activating a redundant communication channel, the system can quickly switch to a backup channel when the main communication channel malfunctions, reacquiring an accurate pressure signal, thereby ensuring the continuity and accuracy of liquid level detection. This mechanism effectively improves the system's fault tolerance and ensures the stable operation of the production process.

[0138] Specifically, pressure sensor transmission anomaly refers to a situation where, during data transmission, the pressure sensor fails to transmit data normally due to various reasons, such as signal loss, data errors, or communication interruption. A redundant communication channel refers to a backup communication line or data transmission method, activated when the main communication channel fails to ensure continuous data transmission. When a transmission anomaly is detected, the system automatically switches to the redundant communication channel to reacquire the pressure signal set. This process includes: after detecting the anomaly, the system issues a switching command to activate the redundant communication channel; reacquires the pressure sensor data through the redundant channel; and preprocesses and analyzes the reacquired data to ensure its accuracy and completeness. Repeating the process from pressure signal acquisition to liquid level determination means the system starts from scratch, including data acquisition, preprocessing, and model calculation steps, to ensure that the final liquid level data is accurate and reliable.

[0139] Preferably, the process of activating the redundant communication channel can be further refined as follows: First, the system monitors the data transmission status of the pressure sensor in real time through a built-in monitoring mechanism. Once a transmission anomaly is detected, such as a data packet loss rate exceeding a set threshold or an incorrect data format, the system immediately triggers the redundant communication mechanism. Specific steps include: the system sends a switching command to the sensor to activate the redundant communication channel; the system reacquires the pressure signal set through the redundant channel, a process that may require recalibrating the sensor or adjusting communication parameters to ensure data accuracy; after acquiring the new pressure signal set, the system repeats data preprocessing, model calculation, and other steps to ensure that the liquid level detection result is not affected by the transmission anomaly. For example, if the main communication channel causes data errors due to electromagnetic interference, the redundant communication channel can avoid the same problem through different transmission paths or encryption methods. Furthermore, the system can be configured with a retry mechanism; if the redundant channel also experiences problems, the system will attempt to switch multiple times until accurate data is obtained. Through this redundancy mechanism, the system can maintain high reliability in complex industrial environments, ensuring the accuracy and continuity of liquid level detection.

[0140] The above embodiments of the present invention have the following beneficial effects:

[0141] 1. By using three sets of pressure sensors—one for the gas phase, one for the slag phase, and one for the copper matte phase—to monitor the liquid level effectively, combined with a dynamic compensation algorithm and a temperature correction coefficient, the measurement error caused by the stratified flow and temperature fluctuations in the molten pool, as described by traditional single-sensor detection, is effectively solved, thus improving the accuracy and reliability of liquid level detection.

[0142] 2. A target liquid level height detection model trained by machine learning is adopted, and combined with interference data removal and redundant communication mechanisms, it overcomes the interference of high-frequency noise from oxygen-enriched side blowing, low-frequency drift from fuel preheating, and abnormal sensor signals on the detection process, thereby enhancing the system's anti-interference capability and fault tolerance.

[0143] 3. By dynamically calculating the slag layer height compensation value, copper matte layer height compensation value, and gas-liquid interaction disturbance correction amount, and generating a visual report containing metallurgical value loss analysis, real-time optimization monitoring of the molten pool liquid level height is realized, providing accurate data support for smelting process adjustment and improving production efficiency and resource utilization.

[0144] like Figure 2 As shown, Figure 2 This is a schematic diagram of the molten pool in a side-blown furnace, a type of equipment commonly used in non-ferrous metal smelting. Its structural design allows gas to be blown in from the side of the furnace to promote mixing and reaction of materials within the molten pool. In a side-blown furnace, the molten pool is divided into several distinct zones, each with its specific function and physical characteristics.

[0145] As shown in the diagram, the side-blown furnace molten pool is mainly divided into three zones: the gas phase zone, the slag phase, and the copper matte phase. The gas phase zone is located at the top of the molten pool and is the area where the gas comes into contact with the molten material. In this zone, oxygen and fuel are injected into the furnace through a spraying system to provide the necessary oxidation and reduction conditions to promote the refining of non-ferrous metals.

[0146] The slag phase, located below the gas phase, is a region composed of molten impurities and flux. The main function of the slag phase is to remove impurities from the raw materials, while its viscosity and density affect its separation efficiency from the copper matte phase.

[0147] The copper matte phase is the bottommost region of the molten pool, mainly composed of copper and other valuable metals. Extraction of the copper matte phase is a primary objective of the side-blown furnace smelting process, and its quality and purity directly affect the value of the final product.

[0148] Multiple pressure sensors (s1, s2, s3) are deployed on the sidewall of the side-blown furnace, corresponding to the copper matte phase, slag phase, and gas phase, respectively. These sensors are used to monitor pressure changes in each region in real time, thereby indirectly reflecting changes in the molten pool level. By analyzing this pressure data, the molten pool level can be accurately calculated, providing crucial information for controlling the smelting process.

[0149] Furthermore, the side-blown lance is a key component of the side-blown furnace, responsible for injecting oxygen-enriched fuel and fuel into the molten pool to promote the reaction. The oxygen-enriched injection system and the fuel injection system provide oxygen and fuel respectively, ensuring the sufficiency and efficiency of the reaction within the furnace.

[0150] In summary, the design and operation of the side-blown furnace molten pool are aimed at maximizing the extraction efficiency of non-ferrous metals while ensuring process safety and economy. A highly efficient smelting process can be achieved by precisely controlling the conditions in different zones within the molten pool.

[0151] The following is for reference. Figure 3The diagram illustrates a structural schematic of an electronic device 300 suitable for implementing some embodiments of the present invention. The electronic devices in some embodiments of the present invention may include, but are not limited to, mobile terminals such as mobile phones, laptops, digital broadcast receivers, PDAs (personal digital assistants), PADs (tablet computers), PMPs (portable multimedia players), in-vehicle terminals (e.g., in-vehicle navigation terminals), and fixed terminals such as digital TVs and desktop computers. Figure 3 The terminal device shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments of the present invention.

[0152] like Figure 3 As shown, the electronic device 300 may include a processing unit (e.g., a central processing unit, a graphics processing unit, etc.) 301, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 302 or a program loaded from a storage device 308 into a random access memory (RAM) 303. The RAM 303 also stores various programs and data required for the operation of the electronic device 300. The processing unit 301, ROM 302, and RAM 303 are interconnected via a bus 304. An input / output (I / O) interface 305 is also connected to the bus 304.

[0153] Typically, the following devices can be connected to I / O interface 305: input devices 306 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; output devices 307 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 308 including, for example, magnetic tapes, hard disks, etc.; and communication devices 309. Communication device 309 allows electronic device 300 to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 3 An electronic device 300 with various devices is shown; however, it should be understood that it is not required to implement or possess all of the devices shown. More or fewer devices may be implemented or possessed alternatively. Figure 3 Each box shown can represent a device or multiple devices as needed.

[0154] Furthermore, the storage medium in the embodiments of this application stores program instructions capable of implementing all the above methods. These program instructions can be stored in the storage medium in the form of a software product, including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) or processor to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks, or terminal devices such as computers, servers, mobile phones, and tablets.

[0155] The above description is merely a selection of preferred embodiments of the present invention and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of the invention as described in the embodiments is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the inventive concept. For example, technical solutions formed by substituting the above-described features with (but not limited to) technical features with similar functions disclosed in the embodiments of the present invention.

Claims

1. A machine learning-based intelligent detection method for liquid level height in a side-blown furnace, characterized in that, include: In response to determining that the detection processing time has been reached, a set of pressure signals is acquired from pressure sensors deployed on the side wall of the side-blown furnace. The pressure sensors include a gas phase pressure sensor located in the gas phase zone, a slag phase pressure sensor located in the slag phase, and a copper matte phase pressure sensor located in the copper matte phase, so as to monitor the pressure changes in the gas phase zone, slag phase, and copper matte phase regions inside the furnace in real time. The pressure signal set is preprocessed to generate a preprocessed pressure signal set; The preprocessed pressure signal set is automatically imported into the associated file corresponding to the target liquid level height detection model and stored in the storage system to generate intelligent detection data of side-blown furnace liquid level height based on machine learning; wherein, the target liquid level height detection model is trained on the sample dataset by machine learning algorithm, and the sample dataset includes pressure data from gas phase pressure sensor, slag phase pressure sensor and copper matte phase pressure sensor and the corresponding actual liquid level height of the molten pool; The height of the molten pool in the side-blown furnace is determined based on the measured liquid level data.

2. The method according to claim 1, characterized in that, The step of preprocessing the pressure signal set to generate a preprocessed pressure signal set includes: An interference data removal operation is performed on the pressure signal set to obtain a preprocessed pressure signal set; wherein, the interference data removal operation includes: Eliminate high-frequency noise caused by the oxygen-enriched side-blowing system from the slag phase pressure sensor signal; Correct the low-frequency drift in the copper matte phase pressure sensor signal caused by the fuel preheating system.

3. The method according to claim 1, characterized in that, The step of automatically importing the preprocessed pressure signal set into the associated file corresponding to the target liquid level detection model and stored in the storage system to generate intelligent detection data for the side-blown furnace liquid level based on machine learning includes: The preprocessed pressure signal set is automatically imported into the associated file corresponding to the target liquid level detection model, so as to generate intelligent detection data of side-blown furnace liquid level based on machine learning by utilizing the data processing script contained in the target liquid level detection model.

4. The method according to claim 1, characterized in that, Determining the height of the molten pool in the side-blown furnace based on the liquid level detection data includes: The first pressure value P measured by the gas phase region pressure sensor in the preprocessed pressure signal set is obtained. s3 The second pressure value P measured by the slag phase pressure sensor s2 The third pressure value P measured by the copper matte phase pressure sensor s1 ; Calculate the molten pool level H using the following formula: Where H is the height of the molten pool; h s h is the dynamic slag layer height compensation value. m ρ is the dynamic copper matte layer height compensation value; Δh is the gas-liquid interaction disturbance correction amount; s ρ is the density of the slag phase; m ρ is the density of the copper matte phase; g is the acceleration due to gravity; D is the inner diameter of the side-blown furnace; α(T) is the temperature compensation coefficient; β(T) is the temperature compensation coefficient; k1 is the turbulence correction coefficient; k2 is the turbulence correction coefficient; P s3 The first pressure value measured by the gas phase pressure sensor; P s2 The second pressure value measured by the slag phase pressure sensor; P s1 This is the third pressure value measured by the copper matte phase pressure sensor.

5. The method according to claim 1, characterized in that, The step of automatically importing the preprocessed pressure signal set into the associated file corresponding to the target liquid level detection model and stored in the storage system to generate intelligent detection data for the side-blown furnace liquid level based on machine learning includes: In response to the total number of pressure sensors being less than a first value, the preprocessed pressure signal set is imported into the associated file of the storage system to generate liquid level detection data; In response to the total number of pressure sensors being higher than the first value, the following operations are performed for each detection area of ​​the side-blown furnace: In response to the number of sensors in the current detection area being higher than the second value, the pressure signal set of the current area is grouped to obtain a pressure signal set, the number of sensors in each group is less than the second value, and the second value is less than the division value, the division value being the first value divided by the number of detection areas. In response to the number of sensors in the current detection area being less than or equal to the second value, the pressure signal set in the current area is determined as a pressure signal set; Determine the total number of pressure signal sets; Assign a unique group identifier to each pressure signal group; Create an association file in the storage system with a number equal to the total number of groups; Set the filenames of all associated files to the corresponding group identifier; Create a number of copies of the target liquid level height detection model in the storage system equal to the total number of groups; Bind each model copy to an associated file with the corresponding filename; Based on the grouping results, the preprocessed pressure signal set is divided into a pressure signal data set and a group identifier data set. Based on the pressure signal data set, group identifier data set, model copy, and associated files, liquid level detection data is generated.

6. The method according to claim 5, characterized in that, The process of generating liquid level detection data based on the pressure signal data set, group identifier data set, model copy, and associated files includes: Perform the following for each pressure signal group: Import the pressure signal data and group identification data of this group into the associated file with the matching filename; The bound model copy is invoked to generate a pressure signal detection data set; By aggregating all pressure signal detection data sets and utilizing the region analysis script contained in the target liquid level detection model, liquid level detection data is generated.

7. The method according to claim 1, characterized in that, The target liquid level detection model supports the use of gradient boosting decision tree algorithm for predicting the liquid level of the molten pool.

8. The method according to claim 1, characterized in that, Also includes: Obtain the liquid level detection confirmation information returned by the side-blown furnace operation terminal; In response to confirmation that the test data is correct, a visual report is generated that includes the side-blown furnace identification, liquid level test data, and liquid level information; The visualization report is sent to the monitoring terminal.

9. The method according to claim 8, characterized in that, The generated visualization report includes: The following procedures were performed on each detection area: gas phase, slag phase, and copper matte phase. Extract the liquid level detection data for the current area; Based on the current regional data, local liquid level height and regional liquid level value information are generated. The regional liquid level value information is calculated using the following formula: V s =η·(h s -h sopt ) 2 ·Q cu V m =μ·(h m -h mopt ) 2 ·Q cu Among them, V s For the metallurgical value loss in the slag phase zone; V m η is the metallurgical value loss in the copper matte zone; η is the metallurgical efficiency coefficient in the slag phase zone, used to measure the impact of the liquid level in the slag phase zone deviating from the optimal value on the metallurgical value; μ is the metallurgical efficiency coefficient in the copper matte zone, used to measure the impact of the liquid level in the copper matte zone deviating from the optimal value on the metallurgical value; h sopt The optimal phase layer height in the slag phase zone; h mopt The optimal phase height for the copper matte region; Q cu For copper matte production; h s h is the dynamic slag layer height compensation value. m This is the dynamic copper matte layer height compensation value.

10. The method according to claim 1, characterized in that, The step of automatically importing the preprocessed pressure signal set into the associated file corresponding to the target liquid level detection model and stored in the storage system to generate intelligent detection data of side-blown furnace liquid level based on machine learning also includes: In response to a pressure sensor transmission anomaly, a redundant communication channel is activated to reacquire the pressure signal set, and the process of acquiring the pressure signal and determining the liquid level is repeated.

Images