A method for measuring the temperature in a tin smelting top-blown furnace and related equipment
Patent Information
- Application Number
- CN202610423538.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-04-01
- Publication Date
- 2026-06-26
Smart Images
Figure CN122286258A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of industrial process monitoring technology, and in particular to a method and related equipment for measuring the temperature inside a top-blown tin smelting furnace. Background Technology
[0002] In the top-blown tin refining process, real-time and accurate monitoring of the furnace temperature is crucial for ensuring product quality and controlling energy consumption. Current technologies primarily rely on thermocouples, infrared temperature measurement devices, or soft sensing methods for temperature monitoring. However, these methods have systemic shortcomings in effectively capturing multi-scale spatiotemporal characteristics and dynamically responding to changes in operating conditions.
[0003] Specifically, traditional thermocouples are typically embedded at a single point on the surface of the furnace tube, only able to obtain temperature readings at local locations. They cannot construct information on the spatial distribution of temperature within the furnace, nor can they capture dynamic changes at different time scales (such as instantaneous fluctuations, medium-term trends, and long-term evolution). Their measurements are essentially isolated point measurements, severely lacking spatiotemporal correlation. Therefore, in environments with strong oxidation, high pressure, and dust, they not only have short lifespans and poor stability, but also cannot achieve continuous sensing of the multi-scale spatiotemporal characteristics of the furnace temperature field.
[0004] While infrared thermometry can acquire high-precision instantaneous temperatures non-contactly, its spatial coverage is limited, making it difficult to penetrate smoke for full-field measurement. It also cannot simultaneously capture temperature evolution across multiple locations and time spans, and lacks the ability to express spatiotemporal coupling characteristics. Furthermore, its hardware fragility and high cost further restrict its sustainable application in complex industrial scenarios.
[0005] Traditional soft sensing methods are mostly based on static modeling, relying on fixed datasets to train a single model. These models struggle to effectively integrate multi-scale time-series information from different sensors and sampling frequencies, failing to fully exploit the spatiotemporal correlations inherent in process variables. Furthermore, when raw material composition changes, operating conditions drift, or equipment status changes slowly, the lack of dynamic adjustment mechanisms prevents the model from adaptively tracking the evolution of the system's spatiotemporal characteristics, leading to a significant decrease in prediction accuracy over time. Especially in many real-world production environments, the scarcity and insufficient spatiotemporal representativeness of labeled data further restrict the accurate modeling of the multi-scale dynamic characteristics of the real temperature field by static models. Summary of the Invention
[0006] This invention provides a method and related equipment for measuring the temperature inside a top-blown furnace in tin smelting, with the aim of improving the accuracy of temperature measurement under conditions where target domain labeled data is extremely scarce.
[0007] To achieve the above objectives, the present invention provides a method for measuring the temperature inside a top-blown tin smelting furnace, comprising: Step 1: Obtain the source domain dataset and the target domain dataset. Both the source domain dataset and the target domain dataset include historical operating data of the top blown furnace in tin smelting. Step 2: Reorganize the one-dimensional time-series data in the source domain dataset to obtain a two-dimensional dynamic matrix; Step 3: Use a two-dimensional dynamic matrix to pre-train the constructed multi-scale parallel convolutional architecture to obtain the first furnace temperature prediction model, and use the target domain dataset to perform knowledge transfer on the first furnace temperature prediction model to obtain the second furnace temperature prediction model. Step 4: The real-time collected tin smelting top blow furnace operating condition data is compared with the historical operating condition data in the target domain dataset to obtain a fine-tuned dataset. The fine-tuned dataset is then used to fine-tune the second furnace temperature prediction model to obtain the third furnace temperature prediction model. Step 5: Input the historical operating data of the target tin smelting top blown furnace into the third furnace temperature measurement model for prediction, and obtain the furnace temperature prediction results for future times. The multi-scale parallel convolutional architecture includes a multi-scale parallel convolutional module, a multi-scale channel attention feature extraction module, a temporal modeling module, and a regression module connected in sequence.
[0008] Furthermore, step 2 includes: Set the sliding window and hysteresis step size; By using a sliding window and hysteresis step size, one-dimensional time-series data in the source domain dataset are reorganized to obtain a two-dimensional dynamic matrix.
[0009] Furthermore, the multi-scale parallel convolution module includes parallel first convolutional layers, second convolutional layers, and third convolutional layers; The inputs of the first, second, and third convolutional layers are all inputs of a multi-scale parallel convolutional architecture, and the outputs of the first, second, and third convolutional layers are all connected to the input of the multi-scale channel attention feature extraction module. The first convolutional layer is used to extract short-range fluctuation features from a two-dimensional dynamic matrix; The second convolutional layer is used to extract mid-range trend features from a two-dimensional dynamic matrix; The third convolutional layer is used to extract long-range evolution features from a two-dimensional dynamic matrix.
[0010] Furthermore, the multi-scale channel attention feature extraction module includes a channel splicing layer, a channel attention layer, a pooling layer, and a flattening layer connected in sequence; The input of the channel splicing layer is connected to the output of the first convolutional layer, the output of the second convolutional layer, and the output of the third convolutional layer, respectively. The output of the flat layer is connected to the input of the timing modeling module.
[0011] Furthermore, step 3 includes: The multi-scale parallel convolutional architecture constructed by inputting a two-dimensional dynamic matrix is used for prediction to obtain the prediction result. The parameters of the multi-scale parallel convolutional architecture are adjusted using the supervision error between the prediction result and the actual furnace temperature to obtain the first furnace temperature prediction model. The source domain dataset and the target domain dataset are respectively input into the network part of the first furnace temperature prediction model for multi-scale spatiotemporal feature extraction and temporal coding to extract high-level features of the source domain dataset and the target domain dataset, thus obtaining source domain features and target domain features. The source domain features and target domain features are input into the gradient inversion layer and then fed into the domain discriminator for domain discriminative adversarial learning. This allows the domain discriminator to distinguish the source domain of features while using the gradient inversion mechanism to inversely constrain the parameter updates of the network part used for multi-scale spatiotemporal feature extraction and temporal coding in the first furnace temperature prediction model. By jointly optimizing the source domain furnace temperature prediction error and the domain discrimination adversarial loss, the parameters of the network part used for multi-scale spatiotemporal feature extraction and temporal coding in the first furnace temperature prediction model are iteratively updated to obtain the second furnace temperature prediction model.
[0012] The present invention also provides a temperature measuring device for a top-blown tin smelting furnace, comprising: The acquisition module is used to acquire source domain datasets and target domain datasets. Both source domain datasets and target domain datasets include historical operating data of the top blown furnace in tin smelting. The recombination module is used to recombine one-dimensional time-series data in the source domain dataset to obtain a two-dimensional dynamic matrix; The training module is used to pre-train the constructed multi-scale parallel convolutional architecture using a two-dimensional dynamic matrix to obtain the first furnace temperature prediction model, and to perform knowledge transfer on the first furnace temperature prediction model using the target domain dataset to obtain the second furnace temperature prediction model. The fine-tuning module is used to measure the similarity between the real-time collected operating data of the top blown furnace of tin smelting and the historical operating data in the target domain dataset to obtain the fine-tuning dataset. The fine-tuning dataset is then used to fine-tune the second furnace temperature prediction model to obtain the third furnace temperature prediction model. The prediction module is used to input the historical operating data of the target tin smelting top blown furnace into the third furnace temperature measurement model for prediction, and obtain the furnace temperature prediction result for future time. The multi-scale parallel convolutional architecture includes a multi-scale parallel convolutional module, a multi-scale channel attention feature extraction module, a temporal modeling module, and a regression module connected in sequence.
[0013] The present invention also provides a terminal device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements a method for measuring the temperature inside a top-blown furnace in tin smelting.
[0014] The present invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements a method for measuring the temperature inside a top-blown furnace in tin smelting.
[0015] The above-described solution of the present invention has the following beneficial effects: This invention acquires source and target domain datasets. One-dimensional time-series data in the source domain dataset is reorganized to obtain a two-dimensional dynamic matrix. This matrix is then pre-trained on a constructed multi-scale parallel convolutional architecture to obtain a first furnace temperature prediction model. Knowledge transfer is performed on the first furnace temperature prediction model using the target domain dataset to obtain a second furnace temperature prediction model. Real-time collected operating data of the top-blown tin smelting furnace is compared with historical operating data in the target domain dataset to obtain a fine-tuning dataset. This fine-tuning dataset is then used to fine-tune the second furnace temperature prediction model to obtain a third furnace temperature prediction model. Finally, historical operating data of the target tin smelting top-blown furnace is input... The temperature is predicted by the third furnace temperature measurement model to obtain the furnace temperature prediction result at future time. The multi-scale parallel convolutional architecture includes a multi-scale parallel convolution module, a multi-scale channel attention feature extraction module, a temporal modeling module, and a regression module connected in sequence. Compared with the prior art, the present invention improves the feature utilization rate through the multi-scale parallel convolutional architecture, uses the target domain dataset to transfer knowledge to the model, and measures the similarity between the real-time collected tin smelting top blown furnace operating condition data and the historical operating condition data in the target domain dataset to obtain a fine-tuning dataset. The model is then fine-tuned using the fine-tuning dataset, which can improve the temperature measurement accuracy under the condition of extremely scarce labeled data in the target domain.
[0016] Other beneficial effects of the present invention will be described in detail in the following detailed description section. Attached Figure Description
[0017] Figure 1 This is a flowchart illustrating an embodiment of the present invention; Figure 2 This is a schematic diagram of the principle framework in an embodiment of the present invention; Figure 3 This is a structural diagram of the temperature measuring device inside the top-blown tin smelting furnace in an embodiment of the present invention; Figure 4 This is a schematic diagram of the structure of the terminal device in an embodiment of the present invention. Detailed Implementation
[0018] To make the technical problems, solutions, and advantages of this invention clearer, a detailed description will be provided below with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.
[0019] In the description of this invention, it should be noted that the terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.
[0020] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a locking connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.
[0021] Furthermore, the technical features involved in the different embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.
[0022] This invention addresses existing problems by providing a method and related equipment for measuring the temperature inside a top-blown tin smelting furnace.
[0023] like Figure 1 , Figure 2 As shown, an embodiment of the present invention provides a method for measuring the temperature inside a top-blown tin smelting furnace, comprising: Step 1: Obtain the source domain dataset and the target domain dataset. Both the source domain dataset and the target domain dataset include historical operating data of the top blown furnace in tin smelting. Step 2: Reorganize the one-dimensional time-series data in the source domain dataset to obtain a two-dimensional dynamic matrix; Step 3: Use a two-dimensional dynamic matrix to pre-train the constructed multi-scale parallel convolutional architecture to obtain the first furnace temperature prediction model, and use the target domain dataset to perform knowledge transfer on the first furnace temperature prediction model to obtain the second furnace temperature prediction model. Step 4: The real-time collected tin smelting top blow furnace operating condition data is compared with the historical operating condition data in the target domain dataset to obtain a fine-tuned dataset. The fine-tuned dataset is then used to fine-tune the second furnace temperature prediction model to obtain the third furnace temperature prediction model. Step 5: Input the historical operating data of the target tin smelting top-blown furnace into the third furnace temperature measurement model for prediction, and obtain the furnace temperature prediction results for future times.
[0024] In this embodiment of the invention, the source domain dataset includes data on coal flow rate, oxygen flow rate of the spray gun, flue gas temperature in the rising flue, and dust collector outlet temperature acquired at multiple time points, as shown in Table 1 below: Table 1 Source Domain Dataset
[0025] In Table 1 above, column 1 represents time points, and columns 2 to 4 represent historical operating data of the top-blown furnace in tin smelting, recorded in chronological order. Traditional methods for analyzing this data typically use a one-dimensional vector at a specific time point without special processing, such as [5480.55, 3770.73, 724.92, 337.02]. Such a data structure is difficult to directly reflect spatiotemporal correlations. To address this issue, this embodiment of the invention uses step 2 to reorganize the one-dimensional time-series data in the source domain dataset to obtain a two-dimensional dynamic matrix.
[0026] Specifically, step 2 includes: Set the sliding window and hysteresis step size; By using a sliding window and hysteresis step size, one-dimensional time-series data in the source domain dataset are reorganized to obtain a two-dimensional dynamic matrix.
[0027] In this embodiment of the invention, the sliding window size is set to 3, and the lag step size is set to 1. Based on the sliding window and the lag step size, the time lag translation technique is used to reorganize the one-dimensional time series data in the source domain dataset to obtain multiple two-dimensional dynamic matrices, as follows: ; ; ; ; As can be seen, the resulting data are all two-dimensional matrices. The data in each row reflects the correlation between the data of different working conditions, and the data in each column reflects the correlation of the data on the time scale. This operation constructs a local spatiotemporal structure by expanding the time dimension, laying the foundation for subsequent feature extraction.
[0028] In actual industrial processes, due to the complex physicochemical reactions and mass transfer mechanisms involved, process variables often exhibit local spatiotemporal patterns. At the same time, these variables are sampled from production equipment and units of different scales, resulting in local behaviors exhibiting multi-scale characteristics. Local behaviors at different scales have different effects on the final output. However, existing soft measurement models are difficult to effectively extract the multi-scale local spatiotemporal characteristics contained in modern multi-coupled industrial processes. Therefore, this invention presents a multi-scale parallel convolutional architecture to process two-dimensional dynamic matrices, including a multi-scale parallel convolutional module, a multi-scale channel attention feature extraction module, a temporal modeling module, and a regression module connected in sequence.
[0029] Specifically, such as Figure 2 As shown, the multi-scale parallel convolution module includes a first convolutional layer, a second convolutional layer, and a third convolutional layer in parallel. The inputs of the first, second, and third convolutional layers are all inputs of a multi-scale parallel convolutional architecture, and the outputs of the first, second, and third convolutional layers are all connected to the input of the multi-scale channel attention feature extraction module. The first convolutional layer is used to extract short-range fluctuation features from a two-dimensional dynamic matrix; The second convolutional layer is used to extract mid-range trend features from a two-dimensional dynamic matrix; The third convolutional layer is used to extract long-range evolution features from a two-dimensional dynamic matrix.
[0030] In this embodiment of the invention, the convolution kernels of the first convolutional layer, the second convolutional layer, and the third convolutional layer are a small convolutional kernel of 3×3, a medium convolutional kernel of 5×5, and a large convolutional kernel of 7×7, respectively, to achieve effective extraction of multi-scale local behavior.
[0031] Specifically, such as Figure 2 As shown, the multi-scale channel attention feature extraction module includes a channel splicing layer, a channel attention layer, a pooling layer, and a flattening layer connected in sequence. The input of the channel splicing layer is connected to the output of the first convolutional layer, the output of the second convolutional layer, and the output of the third convolutional layer, respectively. The output of the flat layer is connected to the input of the time series modeling module, and is used to pass the output single-window feature vector to the time series modeling module.
[0032] In this embodiment of the invention, a channel attention layer is introduced into the multi-scale channel attention feature extraction module to perform nonlinear fusion and adaptive weighting of multi-scale features, so as to strengthen the patterns that make key contributions to temperature prediction and suppress the interference of redundant information, thereby accurately characterizing the different effects of multi-scale local behavior on the output.
[0033] In this embodiment of the invention, the time-series modeling module is a long short-term memory network, which is used to input multiple fused single-window feature vectors into the long short-term memory network to model the time evolution law and long-term dependency relationship of process variables. Through the regression module, time-series recursive calculation is performed to obtain the future multi-step furnace temperature prediction sequence, realizing the end-to-end mapping from real-time sensor data to temperature indicators.
[0034] Specifically, step 3 includes: The multi-scale parallel convolutional architecture constructed by inputting a two-dimensional dynamic matrix is used for prediction to obtain the prediction result. The parameters of the multi-scale parallel convolutional architecture are adjusted using the supervision error between the prediction result and the actual furnace temperature to obtain the first furnace temperature prediction model. The source domain dataset and the target domain dataset are respectively input into the network part of the first furnace temperature prediction model for multi-scale spatiotemporal feature extraction and temporal coding. High-level features of the source domain dataset and the target domain dataset are extracted to obtain source domain features and target domain features. The network part for multi-scale spatiotemporal feature extraction and temporal coding consists of multi-scale parallel convolution, attention feature fusion, and a subsequent temporal modeling network. In this embodiment of the invention, the network part is used as a shared feature extraction module for the source and target domains. Figure 2 As shown; The source domain features and target domain features are input into the gradient inversion layer and then fed into the domain discriminator for domain discriminative adversarial learning. This allows the domain discriminator to distinguish the source domain of features while using the gradient inversion mechanism to inversely constrain the parameter updates of the network part used for multi-scale spatiotemporal feature extraction and temporal coding in the first furnace temperature prediction model. By jointly optimizing the source domain furnace temperature prediction error and the domain discrimination adversarial loss, the parameters of the network part used for multi-scale spatiotemporal feature extraction and temporal coding in the first furnace temperature prediction model are iteratively updated to obtain the second furnace temperature prediction model.
[0035] The source domain features and target domain features are input into the gradient inversion layer and then fed into the domain discriminator for domain discriminative adversarial learning. This allows the domain discriminator to distinguish the source domain of features while using the gradient inversion mechanism to inversely constrain the parameter updates of the feature extraction module. By jointly optimizing the source domain furnace temperature prediction error and the domain discrimination adversarial loss, the parameters of the feature extraction module are iteratively updated to obtain the second furnace temperature prediction model.
[0036] To overcome the bottleneck of scarce target domain datasets, this invention integrates a domain adversarial training mechanism during model training. After the source and target domain datasets are processed by the feature extraction module, the adversarial learning of the gradient reversal layer and the domain discriminator forces the feature space to eliminate inter-domain distribution differences, thereby achieving knowledge transfer under unsupervised conditions.
[0037] To address the persistent operational condition drift problem in industrial scenarios, this invention measures the similarity between real-time collected operating condition data of the top-blown furnace in tin smelting and historical operating condition data in the target domain dataset to obtain a fine-tuned dataset. This fine-tuned dataset is then used to perform lightweight backpropagation iteration on the second furnace temperature prediction model, enabling rapid online updates of model parameters and ensuring long-term prediction robustness.
[0038] The present invention further illustrates the above-mentioned furnace temperature measurement method in a specific scenario, and the specific process is as follows: First, based on the collected experimental data of the top-blown furnace operation, a systematic analysis of the statistical correlation between various characteristic variables and furnace temperature was conducted. Specifically, correlation coefficients were used to calculate the correlation coefficients of various operating conditions, such as coal flow rate, lance oxygen flow rate, electrostatic precipitator outlet temperature, bag filter outlet temperature, waste heat boiler convection zone flue gas temperature, rising flue gas temperature, top-blown furnace cooling water flow rate, and boiler steam volume, to construct a complete correlation coefficient matrix. The correlation coefficient matrix can characterize that the stronger the combustion, the higher the flue gas temperature, and the greater the steam and load, the higher the overall furnace temperature tends to rise, while the increase in cooling water slightly inhibits the rise in furnace temperature. The overall correlation direction is consistent with the physical mechanism of the smelting furnace, but due to the influence of multi-factor coupling and the control system, the linear correlation of a single variable with the furnace temperature is not particularly strong. Next, the experimental data above was used to construct a sliding time window of length 10 in chronological order to obtain experimental samples, of which the first 80% was used for training and the last 20% for testing. The evaluation results are shown in Table 2 below: Table 2 Model Evaluation Results
[0039] As shown in Table 2 above, after training, the RMSE on the training set is approximately 4.2℃, the MAE is approximately 3.3℃, and the coefficient of determination R² reaches 0.943; on the test set, the RMSE is approximately 5.3℃, the MAE is approximately 4.3℃, and the R² is 0.917. The model can already characterize the nonlinear mapping relationship between furnace temperature and multiple process characteristic variables well, and the prediction error is controlled at around 5℃, which meets the engineering application requirements for process monitoring and operation guidance.
[0040] This invention acquires a source domain dataset and a target domain dataset. One-dimensional time-series data in the source domain dataset is reorganized to obtain a two-dimensional dynamic matrix. A multi-scale parallel convolutional architecture is pre-trained to obtain a first furnace temperature prediction model. Knowledge transfer is then performed on the first furnace temperature prediction model using the target domain dataset to obtain a second furnace temperature prediction model. Real-time collected operating data of the top-blown tin smelting furnace is compared with historical operating data in the target domain dataset to obtain a fine-tuning dataset. This fine-tuning dataset is then used to fine-tune the second furnace temperature prediction model to obtain a third furnace temperature prediction model. Finally, historical operating data of the target tin smelting top-blown furnace is input... The temperature is predicted by the third furnace temperature measurement model to obtain the furnace temperature prediction result at future time. The multi-scale parallel convolutional architecture includes a multi-scale parallel convolution module, a multi-scale channel attention feature extraction module, a temporal modeling module, and a regression module connected in sequence. Compared with the prior art, the embodiment of the present invention improves the feature utilization rate through the multi-scale parallel convolutional architecture, uses the target domain dataset to perform knowledge transfer on the model, and measures the similarity between the real-time collected tin smelting top blown furnace operating condition data and the historical operating condition data in the target domain dataset to obtain a fine-tuning dataset. The model is fine-tuned using the fine-tuning dataset, which can improve the temperature measurement accuracy under the condition of extremely scarce labeled data in the target domain.
[0041] Corresponding to the temperature measurement method inside the top-blown furnace for tin smelting described in the above embodiments, such as Figure 3 As shown, this embodiment of the invention also provides a temperature measuring device 100 for a top-blown tin smelting furnace, which includes: The acquisition module 101 is used to acquire source domain datasets and target domain datasets. Both source domain datasets and target domain datasets include historical operating data of the top blown furnace in tin smelting. The recombination module 102 is used to recombine the one-dimensional time-series data in the source domain dataset to obtain a two-dimensional dynamic matrix; Training module 103 is used to pre-train the constructed multi-scale parallel convolutional architecture using a two-dimensional dynamic matrix to obtain a first furnace temperature prediction model, and to perform knowledge transfer on the first furnace temperature prediction model using the target domain dataset to obtain a second furnace temperature prediction model. The fine-tuning module 104 is used to measure the similarity between the real-time collected tin smelting top blow furnace operating condition data and the historical operating condition data in the target domain dataset to obtain the fine-tuning dataset, and to fine-tune the second furnace temperature prediction model using the fine-tuning dataset to obtain the third furnace temperature prediction model. The prediction module 105 is used to input the historical operating data of the target tin smelting top blown furnace into the third furnace temperature measurement model for prediction, and obtain the furnace temperature prediction result for future time. The multi-scale parallel convolutional architecture includes a multi-scale parallel convolutional module, a multi-scale channel attention feature extraction module, a temporal modeling module, and a regression module connected in sequence.
[0042] It should be noted that the information interaction and execution process between the above-mentioned devices / units are based on the same concept as the method embodiments of this application. For details on their specific functions and technical effects, please refer to the method embodiments section, and they will not be repeated here.
[0043] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this application. The specific working process of the units and modules in the above system can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.
[0044] This invention also provides a terminal device, such as... Figure 4 As shown, the terminal device D10 of this embodiment includes: at least one processor D100 ( Figure 4 The diagram shows only one processor, a memory D101, and a computer program D102 stored in the memory D101 and executable on the at least one processor D100. When the processor D100 executes the computer program D102, it implements the above-described method for measuring the temperature inside a top-blown furnace in tin smelting.
[0045] The terminal device D10 can be a desktop computer, laptop, handheld computer, server, server cluster, or cloud server, etc. This terminal device may include, but is not limited to, a processor D100 and a memory D101. Those skilled in the art will understand that... Figure 4 This is merely an example of terminal device D10 and does not constitute a limitation on terminal device D10. It may include more or fewer components than shown in the figure, or combine certain components, or different components, such as input / output devices, network access devices, etc.
[0046] The processor D100 can be a central processing unit (CPU), or it can be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor.
[0047] In some embodiments, the memory D101 may be an internal storage unit of the terminal device D10, such as a hard disk or memory of the terminal device D10. In other embodiments, the memory D101 may be an external storage device of the terminal device D10, such as a plug-in hard disk, smart media card (SMC), secure digital card (SD), flash card, etc., equipped on the terminal device D10. Furthermore, the memory D101 may include both internal and external storage units of the terminal device D10. The memory D101 is used to store the operating system, applications, bootloader, data, and other programs, such as the program code of the computer program. The memory D101 can also be used to temporarily store data that has been output or will be output.
[0048] It should be noted that the information interaction and execution process between the above-mentioned devices / units are based on the same concept as the method embodiments of this application. For details on their specific functions and technical effects, please refer to the method embodiments section, and they will not be repeated here.
[0049] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this application. The specific working process of the units and modules in the above system can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.
[0050] This invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements a method for measuring the temperature inside a top-blown tin smelting furnace.
[0051] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of this application can be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include at least: any entity or device capable of carrying the computer program code to a building device / terminal device, a recording medium, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium. Examples include USB flash drives, portable hard drives, magnetic disks, or optical disks.
[0052] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.
Claims
1. A method for measuring the temperature in a tin smelting top blown converter, characterized in that include: Step 1: Obtain the source domain dataset and the target domain dataset, both of which include historical operating data of the top blown furnace in tin smelting. Step 2: Reorganize the one-dimensional time-series data in the source domain dataset to obtain a two-dimensional dynamic matrix; Step 3: Use the two-dimensional dynamic matrix to pre-train the constructed multi-scale parallel convolutional architecture to obtain the first furnace temperature prediction model, and use the target domain dataset to perform knowledge transfer on the first furnace temperature prediction model to obtain the second furnace temperature prediction model. Step 4: Perform a similarity measurement between the real-time collected tin smelting top blown furnace operating condition data and the historical operating condition data in the target domain dataset to obtain a fine-tuned dataset. Then, use the fine-tuned dataset to fine-tune the second furnace temperature prediction model to obtain the third furnace temperature prediction model. Step 5: Input the historical operating data of the target tin smelting top blown furnace into the third furnace temperature measurement model for prediction, and obtain the furnace temperature prediction result for future time. The multi-scale parallel convolutional architecture includes a multi-scale parallel convolutional module, a multi-scale channel attention feature extraction module, a temporal modeling module, and a regression module connected in sequence.
2. The tin smelting top blown converter furnace temperature measurement method according to claim 1, characterized in that, Step 2 includes: Set the sliding window and hysteresis step size; The one-dimensional time-series data in the source domain dataset is reorganized using the sliding window and hysteresis step size to obtain a two-dimensional dynamic matrix.
3. The method of claim 1, wherein the temperature of the tin smelting top- blown furnace is measured by a method comprising: The multi-scale parallel convolution module includes a first convolutional layer, a second convolutional layer, and a third convolutional layer in parallel. The input terminals of the first convolutional layer, the second convolutional layer, and the third convolutional layer are all input terminals of the multi-scale parallel convolutional architecture, and the output terminals of the first convolutional layer, the second convolutional layer, and the third convolutional layer are all connected to the input terminal of the multi-scale channel attention feature extraction module. The first convolutional layer is used to extract short-range fluctuation features from the two-dimensional dynamic matrix; The second convolutional layer is used to extract mid-range trend features from the two-dimensional dynamic matrix; The third convolutional layer is used to extract long-range evolution features from the two-dimensional dynamic matrix.
4. The method of claim 3, wherein the temperature of the molten tin is measured by a method comprising: The multi-scale channel attention feature extraction module includes a channel splicing layer, a channel attention layer, a pooling layer, and a flattening layer connected in sequence. The input terminals of the channel splicing layer are respectively connected to the output terminals of the first convolutional layer, the second convolutional layer, and the third convolutional layer; The output of the flat layer is connected to the input of the timing modeling module.
5. The method of claim 1, wherein the temperature of the tin smelting top- blown furnace is measured by a method comprising: Step 3 includes: The multi-scale parallel convolutional architecture constructed by inputting the two-dimensional dynamic matrix is used to make predictions, and the prediction results are obtained. The parameters of the multi-scale parallel convolutional architecture are adjusted using the supervision error between the prediction results and the actual furnace temperature to obtain the first furnace temperature prediction model. The source domain dataset and the target domain dataset are respectively input into the network part of the first furnace temperature prediction model for multi-scale spatiotemporal feature extraction and temporal coding to extract high-level features of the source domain dataset and the target domain dataset, thereby obtaining source domain features and target domain features. The source domain features and the target domain features are input together into the gradient inversion layer and then further fed into the domain discriminator for domain discriminative adversarial learning. This allows the domain discriminator to distinguish the source domain of the features while using the gradient inversion mechanism to inversely constrain the parameter updates of the network part used for multi-scale spatiotemporal feature extraction and temporal coding in the first furnace temperature prediction model. By jointly optimizing the source domain furnace temperature prediction error and the domain discrimination adversarial loss, the parameters of the network part used for multi-scale spatiotemporal feature extraction and temporal coding in the first furnace temperature prediction model are iteratively updated to obtain the second furnace temperature prediction model.
6. A tin smelting top blown vessel in situ temperature measuring device characterised in that, include: The acquisition module is used to acquire source domain datasets and target domain datasets, both of which include historical operating data of the top blown furnace in tin smelting. The recombination module is used to recombine the one-dimensional time-series data in the source domain dataset to obtain a two-dimensional dynamic matrix; The training module is used to pre-train the constructed multi-scale parallel convolutional architecture using the two-dimensional dynamic matrix to obtain a first furnace temperature prediction model, and to perform knowledge transfer on the first furnace temperature prediction model using the target domain dataset to obtain a second furnace temperature prediction model. The fine-tuning module is used to measure the similarity between the real-time collected operating data of the top blown furnace of tin smelting and the historical operating data in the target domain dataset to obtain the fine-tuning dataset, and to fine-tune the second furnace temperature prediction model using the fine-tuning dataset to obtain the third furnace temperature prediction model. The prediction module is used to input the historical operating data of the target tin smelting top blown furnace into the third furnace temperature measurement model for prediction, and obtain the furnace temperature prediction result for future time. The multi-scale parallel convolutional architecture includes a multi-scale parallel convolutional module, a multi-scale channel attention feature extraction module, a temporal modeling module, and a regression module connected in sequence.
7. A terminal device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the method for measuring the temperature inside a top-blown tin smelting furnace as described in any one of claims 1 to 5.
8. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the method for measuring the temperature inside a top-blown tin smelting furnace as described in any one of claims 1 to 5.