Automatic control method and system for fly ash melting processing system

Abstract

The present application relates to the field of intelligent control, and provides an automatic control method and system for a fly ash melting processing system, which uses an image analysis and data processing algorithm based on artificial intelligence to perform state analysis of a molten liquid state image and time series correlation of the melting temperature data in a melting furnace, thereby intelligently obtaining a recommended decoding value of the heating power at the next time point based on the prompt enhanced joint representation between the molten liquid state characteristics and the melting temperature time series correlation characteristics. In this way, compared with the traditional fixed temperature setting value, the system can analyze the state of the melting process in real time, and dynamically and accurately adjust the heating power to ensure that the temperature is always kept within the optimal range, thereby reducing temperature fluctuations, maintaining the stability of the melting process, and realizing intelligent monitoring and control of the fly ash melting process.

Description

Technical Field

[0001] The present application relates to the field of intelligent control, and more specifically, to an automatic control method and system for a fly ash melting processing system. Background Art

[0002] Fly ash mainly comes from industrial facilities such as coal-fired power plants and waste incineration plants, and contains a large amount of fine particles and harmful substances, such as heavy metals and dioxins. These harmful substances not only pollute the environment, but may also have serious effects on human health. Fly ash melting treatment is an effective waste treatment technology that converts fly ash into stable glass products through high-temperature melting, thereby achieving harmlessness and resource utilization of waste. This process can not only reduce harmful substances in fly ash, but also convert it into valuable building materials.

[0003] In the fly ash melting process, the high-temperature melting stage is the core part of the entire process. Its main function is to melt the fly ash into a liquid at high temperature, and then form a stable glass product through subsequent water quenching. In this process, temperature control is particularly important. However, the traditional fly ash melting treatment system adopts a fixed temperature setting value, which cannot be dynamically adjusted according to the actual working conditions. This static control method results in insufficient precision of temperature control and is difficult to adapt to the complex melting process. Especially when the fly ash composition and feed rate change, the fixed temperature setting value cannot effectively cope with it, which may lead to insufficient melting or overheating. In addition, the traditional temperature control system has a slow response speed. When the temperature in the melting furnace changes rapidly, the system may not be able to make adjustments in time, resulting in large temperature fluctuations. This hysteresis phenomenon not only affects the stability of the melting process, but also may increase energy consumption and thermal stress of equipment, and accelerate equipment aging.

[0004] Therefore, an automatic control solution for a fly ash melting processing system is desired. Summary of the invention

[0005] In view of the shortcomings of the prior art, the present application provides an automatic control method and system for a fly ash melting processing system.

[0006] According to one aspect of the present application, an automatic control method for a fly ash melting treatment system is provided, and the fly ash melting treatment process includes the following steps: washing the fly ash to be treated with water to obtain washed fly ash; drying the washed fly ash to obtain dried washed fly ash; mixing the dried washed fly ash with solid waste to form a shape, and then transporting it together with fuel carbon to a melting furnace for high-temperature melting treatment to form a molten liquid and molten waste gas; quenching the molten liquid to form a glass product; screening and packaging the glass product; wherein the automatic control method for the fly ash melting treatment system is applied to the steps of: mixing the dried washed fly ash with solid waste to form a shape, and then transporting it together with fuel carbon to a melting furnace for high-temperature melting treatment to form a molten liquid and a molten waste gas. in the exhaust gas; wherein, the automatic control method of the fly ash melting processing system comprises: obtaining a molten liquid state image collected by a high-temperature camera deployed at the observation window of the melting furnace, and obtaining a time series of melting temperature data in the melting furnace; performing state feature extraction on the molten liquid state image to obtain a molten liquid state feature; performing time series feature extraction on the time series of the melting temperature data in the melting furnace to obtain a melting temperature time series correlation feature; performing state-temperature prompt enhancement on the molten liquid state feature and the melting temperature time series correlation feature to obtain a molten state-temperature cross-modal guidance joint coding feature; based on the molten state-temperature cross-modal guidance joint coding feature, obtaining a control result, and the control result is used to represent a recommended decoding value of the heating power at the next time point.

[0007] According to another aspect of the present application, an automatic control system for a fly ash melting processing system is provided, which is used to execute the automatic control method for the fly ash melting processing system as described above, wherein the automatic control system for the fly ash melting processing system comprises: a data acquisition module, which is used to obtain a molten liquid state image collected by a high-temperature camera deployed at the observation window of the melting furnace, and obtain a time series of melting temperature data in the melting furnace; a molten liquid state feature extraction module, which is used to extract state features of the molten liquid state image to obtain molten liquid state features; a molten temperature time series feature extraction module, which is used to extract time series features of the time series of molten temperature data in the melting furnace to obtain molten temperature time series correlation features; a molten state-temperature cross-modal joint encoding module, which is used to perform state-temperature prompt enhancement on the molten liquid state features and the molten temperature time series correlation features to obtain a molten state-temperature cross-modal guidance joint encoding feature; a heating power recommendation result generation module, which is used to obtain a control result based on the molten state-temperature cross-modal guidance joint encoding feature, and the control result is used to represent the recommended decoding value of the heating power at the next time point.

[0008] The present application has significant technical effects due to the adoption of the above technical solutions: the automatic control method and system of the fly ash melting processing system provided by the present application adopts an image analysis and data processing algorithm based on artificial intelligence to perform state analysis of the molten liquid state image and time series correlation of the melting temperature data in the melting furnace, so as to intelligently obtain the recommended decoding value of the heating power at the next time point according to the prompt enhanced joint representation between the molten liquid state characteristics and the melting temperature time series correlation characteristics. In this way, compared with the traditional fixed temperature setting value, the system can analyze the state of the melting process in real time, and dynamically and accurately adjust the heating power to ensure that the temperature is always kept within the optimal range, thereby reducing temperature fluctuations, maintaining the stability of the melting process, and realizing intelligent monitoring and control of the fly ash melting process. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0010] Figure 1 Flow chart of the water washing treatment steps in the fly ash melting process.

[0011] Figure 2 The figure is a flow chart of the high temperature melting treatment and water quenching treatment steps in the fly ash melting process.

[0012] Figure 3 The present invention is a flowchart of an automatic control method of a fly ash melting processing system according to an embodiment of the present application.

[0013] Figure 4 This is a flow chart of step S140 in the automatic control method of the fly ash melting processing system according to an embodiment of the present application.

[0014] Figure 5 This is a flow chart of step S142 in the automatic control method of the fly ash melting processing system according to an embodiment of the present application.

[0015] Figure 6 It is a block diagram of an automatic control system of a fly ash melting processing system according to an embodiment of the present application. DETAILED DESCRIPTION

[0016] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0017] It should be understood that the fly ash melting process includes the following steps: washing the fly ash to be treated with water to obtain washed fly ash; drying the washed fly ash to obtain dried washed fly ash; mixing the dried washed fly ash with solid waste to form a high-temperature melting treatment and then transporting it together with fuel carbon to a melting furnace to form a molten liquid and molten waste gas; quenching the molten liquid with water to form a glassy product; and screening and packaging the glassy product.

[0018] Specifically, the fly ash needs to be pre-treated by washing to remove chloride salts and other soluble harmful substances. By washing, the corrosive problems in the subsequent treatment process can be reduced, and the chlorine content in the melting process can be reduced, which can reduce the corrosion of equipment and the difficulty of flue gas treatment. Then, considering that the washed fly ash contains a high moisture content, it is necessary to remove the excess moisture by drying to facilitate the subsequent melting process. Here, air drying or mechanical drying is usually used to ensure that the water content of the fly ash meets the requirements of the melting process. Then the dried fly ash is mixed with solid waste (such as clay, cement, etc.) to adjust its physical and chemical properties to make it more suitable for melting. The purpose of mixing and forming is to improve the melting characteristics of the fly ash, increase its stability, and help reduce energy consumption in the melting process. The mixed and formed fly ash is transported to the melting furnace together with the fuel carbon and melted at a high temperature of more than 1200°C. In this process, harmful substances in the fly ash, such as dioxins and heavy metals, are decomposed or physically encapsulated at high temperatures and converted into a molten state. The melted fly ash needs to be water quenched to quickly reduce its temperature and form a glassy product. Water quenching not only helps to form a stable glassy structure, but also effectively fixes heavy metals and reduces their leaching risk. Finally, the formed glassy product needs to be screened to remove large impurities and incompletely melted particles. The screened glassy product can be packaged for safe storage and transportation.

[0019] More specifically, Figure 1 The following is a flow chart of the water washing process in the fly ash melting process. Figure 1As shown, the fly ash melting process starts with the fly ash transport vehicle, and the fly ash is transported to the treatment facility and stored in the raw ash silo. Subsequently, the fly ash is sent to the pulping tank through pneumatic conveying and mechanical conveying, and mixed with water to form a slurry. In the pulping tank, the fly ash slurry undergoes preliminary mechanical stirring to prepare for subsequent water washing and filtration. Next, the fly ash slurry enters the primary water washing tank to remove some pollutants through water washing. The filtrate generated during the water washing process is sent to the wastewater treatment facility for treatment. The fly ash slurry that has undergone primary water washing is transferred to the filter through a slurry pump to further remove pollutants and moisture. The filtered fly ash slurry enters the secondary water washing tank again for secondary water washing to further reduce the pollutant content. In the tertiary water washing tank, tap water and stabilizers may be added to adjust the chemical properties of the fly ash slurry to make it more suitable for the subsequent high-temperature melting process. The fly ash slurry is washed for the third time in the tertiary water washing tank to ensure that the pollutant content meets the safety standard. The fly ash slurry that has undergone tertiary water washing is transferred to the fly ash drying equipment through a conveyor belt to further remove moisture and form dry fly ash. Finally, the dried fly ash is sent to a high-temperature melting furnace. Under high temperature, harmful substances in the fly ash are solidified in the molten glassy material. Throughout the process, wastewater treatment facilities ensure that the wastewater generated during the water washing process is properly treated to avoid secondary pollution. It is worth noting that in the water washing step, the filtrate produced by the tertiary water washing can be reused in the secondary water washing tank, and the filtrate produced by the secondary water washing can be reused in the primary water washing tank. The water washing liquid used in the tertiary water washing tank is supplemented by tap water, and the water washing wastewater generated by the primary water washing enters the water washing wastewater treatment system through a pipe. And the water washing water in the water washing step can use the wastewater of petrochemical enterprises in the park. Through this treatment, waste treatment and coordinated disposal can be achieved in the water washing process.

[0020] More specifically, Figure 2 The figure is a flow chart of the high temperature melting treatment and water quenching treatment steps in the fly ash melting process. Figure 2As shown, first, raw materials such as dried fly ash, silicon slag, carbonized slag, surface treatment waste, waste alkali, hazardous incineration slag, domestic waste incineration slag, silicon powder waste, foundry industry shot blasting sand, carbon concentrate, etc. are collected and pre-treated, including classification, crushing, screening and magnetic separation, to improve the efficiency of subsequent treatment. After pretreatment, the raw materials are sent to a high-temperature melting furnace and melted at a high temperature of more than 1200°C to fix the harmful substances in the molten liquid. The molten material is rapidly cooled by water quenching to form a vitreous product, which can be used for hull sandblasting, low-temperature foaming insulation flame retardant materials, coating additives, etc. The flue gas generated during the high-temperature melting process needs to go through a series of treatment steps to remove harmful gases and particulate matter. The flue gas is first removed from nitrogen oxides (NOx) by selective catalytic reduction (SCR) technology, and then large particles are removed by gravity pre-dust removal. Then, the flue gas is cooled by surface cooling to prepare for subsequent dust removal and desulfurization. Activated carbon adsorption is used to remove harmful gases in the flue gas, such as dioxins and furans. After that, the bag filter further removes fine particles. Sodium desulfurization and deacidification technology is used to remove sulfides and acid gases in the flue gas. Finally, white emission elimination technology reduces the emission of visible white smoke, that is, water vapor. During the entire treatment process, some valuable by-products may be produced, such as metal recovery, specific chemicals, etc., which can be further recycled. In this way, through this series of delicate steps, a variety of industrial and urban wastes can be effectively treated in the high-temperature melting treatment step, reducing environmental pollution and realizing the recycling of resources.

[0021] In the fly ash melting process, the high-temperature melting stage is the key link, the purpose of which is to convert the fly ash into liquid and eventually form a stable glass. Temperature control is crucial to this process, but the traditional processing system uses static temperature settings and lacks the ability to dynamically respond to changes in operating conditions. This control method is not flexible enough when the fly ash composition and feed rate change, which may lead to uneven melting or excessive temperature. In addition, the traditional temperature control system has a slow response speed and is difficult to respond to rapid temperature changes in time, which can easily cause temperature fluctuations, thereby affecting melting stability and increasing energy consumption, and may aggravate thermal stress and aging of equipment.

[0022] Based on this, the present application proposes an automatic control method for a fly ash melting treatment system, which will be applied to the step of "combining the dried washed fly ash with solid waste to form a mixture and then transporting it to a melting furnace together with fuel carbon to perform high-temperature melting treatment to form a molten liquid and molten waste gas". Accordingly, the technical concept of the present application is to obtain a molten liquid state image collected by a high-temperature camera deployed at the observation window of the melting furnace, and to obtain a time series of the melting temperature data in the melting furnace, and to use an image analysis and data processing algorithm based on artificial intelligence to perform state analysis of the molten liquid state image and time series association of the melting temperature data in the melting furnace, so as to intelligently obtain the recommended decoding value of the heating power at the next time point according to the prompt enhanced joint representation between the molten liquid state characteristics and the melting temperature time series association characteristics. In this way, compared with the traditional fixed temperature setting value, the system can analyze the state of the melting process in real time, and dynamically and accurately adjust the heating power to ensure that the temperature is always kept within the optimal range, thereby reducing temperature fluctuations and maintaining the stability of the melting process, so as to realize intelligent monitoring and control of the fly ash melting process.

[0023] Figure 3 FIG. 1 is a flow chart of an automatic control method for a fly ash melting processing system according to an embodiment of the present application. Figure 3 As shown, the automatic control method of the fly ash melting processing system according to the embodiment of the present application includes: S110, obtaining a molten liquid state image collected by a high-temperature camera deployed at the observation window of the melting furnace, and obtaining a time series of melting temperature data in the melting furnace; S120, performing state feature extraction on the molten liquid state image to obtain a molten liquid state feature; S130, performing time series feature extraction on the time series of the melting temperature data in the melting furnace to obtain a melting temperature time series correlation feature; S140, performing state-temperature prompt enhancement on the molten liquid state feature and the melting temperature time series correlation feature to obtain a molten state-temperature cross-modal guidance joint coding feature; S150, obtaining a control result based on the molten state-temperature cross-modal guidance joint coding feature, and the control result is used to represent a recommended decoding value of the heating power at the next time point.

[0024] In step S110, a molten liquid state image collected by a high-temperature camera deployed at the observation window of the melting furnace is obtained, and a time series of the melting temperature data in the melting furnace is obtained. It should be understood that the molten liquid state image mainly reflects the real-time state of the molten liquid (i.e., a liquid mixture formed by fly ash, solid waste, and fuel carbon at high temperature) in the melting furnace. From the image, the color and brightness information of the molten liquid can be understood, which can indirectly reflect the degree of melting of the molten liquid, such as the molten liquid gradually changes from dark red to bright yellow as the temperature increases. In addition, the fluidity information, bubble and impurity information of the molten liquid can also be understood. In addition to the melting temperature information at different time points, the time series of the melting temperature data also reveals the changing trend and fluctuation information of the melting temperature. In general, the molten liquid state image provides real-time feedback of the melting process, which can reflect whether the current state of the molten liquid meets expectations; and the time series of the temperature data can help analyze the temperature trend and fluctuation in the melting process. By fusing and analyzing these two types of data through a deep learning model, the correlation characteristics between the molten liquid state and temperature can be extracted, and the heating power can be adjusted accordingly. In this way, the melting effect is optimized by dynamically adjusting the heating power to ensure efficient and stable operation of the fly ash melting processing system.

[0025] In the fly ash melting treatment system, in order to achieve precise control of the melting process, it is necessary to obtain real-time images of the state of the molten liquid in the melting furnace and the time series of the melting temperature data. This process involves multiple links such as the installation and image acquisition of high-temperature cameras, the arrangement and data collection of temperature sensors, etc. Each link needs to be carefully designed and implemented to ensure the accuracy and reliability of the data.

[0026] First, a high-temperature camera needs to be deployed at the observation window of the melting furnace. The melting furnace usually works in an extremely high temperature environment, up to 1200°C or more, which places extremely high demands on the camera's high-temperature resistance. Therefore, the selected high-temperature camera must have good high-temperature resistance and protection measures to ensure that it can work normally in a high-temperature environment. Specifically, high-temperature cameras usually use special lens and sensor materials that can withstand high-temperature environments without damage. For example, the camera lens may be made of quartz glass, which has extremely high heat resistance and transparency, and can maintain a clear field of view at high temperatures. In addition, the camera also needs to be equipped with an effective heat dissipation device, such as a water cooling or air cooling system, to prevent overheating damage caused by long-term high-temperature operation. During the installation process, the location selection of the high-temperature camera is also very important. Usually, the best viewing angle of the melting furnace observation window is selected to ensure that the camera can clearly capture the surface state of the molten liquid, including key features such as color, brightness, and bubble distribution. In order to ensure image quality, the camera also needs to be cleaned and maintained regularly to prevent dust and residue from affecting the lens.

[0027] Afterwards, the molten liquid state images collected by the high-temperature camera will be transmitted to the control system in real time through a high-speed data transmission line. These image data will be stored in a high-performance image processing server for subsequent feature extraction and analysis.

[0028] In step S120, the state feature extraction of the molten liquid state image is performed to obtain the molten liquid state feature. Specifically, in the embodiment of the present application, the step S120 includes: passing the molten liquid state image through a molten liquid state feature extractor based on a deep model to obtain a molten liquid state feature map as the molten liquid state feature. The molten liquid state feature extractor based on the deep model described here is a molten liquid state feature extractor based on a deep separable convolutional neural network model. It should be understood that considering that the molten liquid state image contains subtle changes in the molten liquid, such as fluidity, color changes, bubble distribution, etc. And considering that the deep separable convolutional neural network model significantly reduces the amount of calculation and the number of parameters by decomposing the traditional convolution operation into deep convolution and point-by-point convolution, at the same time, through different levels of convolution operations, it can extract the characteristics of the molten liquid at different scales, such as texture, shape, color, etc., these features are crucial for judging the state of the melting process. Based on this, in the technical solution of the present application, the molten liquid state image is passed through a molten liquid state feature extractor based on a deep separable convolutional neural network model to capture and extract feature information of different scales and levels of the molten liquid state, thereby obtaining a molten liquid state feature map. In particular, the deep separable convolutional neural network model effectively reduces the computational complexity by decomposing the standard convolution operation into two independent steps, namely, deep convolution and point-by-point convolution. Specifically, in deep convolution, for each input channel in the molten liquid state image, such as the RGB three-color channel, a single small convolution kernel (usually a 3x3 convolution kernel) is applied, and this convolution kernel only operates on the data of the channel. In this way, each input channel generates an output channel after convolution, retaining the original spatial structure information while reducing the amount of calculation. Subsequently, through a 1x1 point-by-point convolution operation, the information of these output channels is cross-channel fused to change the number of output channels, which effectively reduces the amount of calculation and the number of parameters, while maintaining effective feature extraction of the molten liquid state image, and finally generates a molten liquid state feature map.

[0029] In step S130, the time series of the melting temperature data in the melting furnace is subjected to time series feature extraction to obtain the melting temperature time series correlation feature. Specifically, in the embodiment of the present application, the step S130 includes: the time series of the melting temperature data in the melting furnace is passed through the melting temperature time series feature extractor based on the forward LSTM model to obtain the melting temperature time series correlation feature vector as the melting temperature time series correlation feature. It should be understood that the temperature change in the melting process is a typical time series problem, and the historical temperature data has an important influence on the current and future temperature. That is, there are different temperature time series dynamic change laws between different time scales in the time series of the melting temperature data in the melting furnace. Therefore, in order to effectively capture the change trend and correlation pattern of the melting temperature between different time spans over time, in the technical solution of the present application, the time series of the melting temperature data in the melting furnace is passed through the melting temperature time series feature extractor based on the forward LSTM model to utilize the characteristics of the forward LSTM model that is good at capturing the long-term dependence of time to mine and capture the temperature time series dynamic correlation in different time periods, and obtain the melting temperature time series correlation feature vector. That is, the forward LSTM model is designed with a special gating mechanism (input gate, forget gate, and output gate) that can learn from past data and update its internal state in real time when processing time series data. This enables the model to monitor the temperature changes of the melting furnace in real time, thereby better understanding and predicting the future trend of melting temperature.

[0030] In step S140, the melt state feature and the melt temperature time series correlation feature are combined with state-temperature prompt enhancement to obtain a melt state-temperature cross-modal guidance joint coding feature. Specifically, Figure 4 FIG. 1 is a flow chart of step S140 in the automatic control method of the fly ash melting processing system according to an embodiment of the present application. Figure 4 As shown, the step S140 includes: S141, performing local feature decomposition along the channel dimension on the molten liquid state feature map to obtain a set of molten liquid state local feature matrices; S142, performing prompt-assisted cross-modal guided combination on the set of the melt temperature time series correlation feature vector and the molten liquid state local feature matrix to obtain the melt state-temperature cross-modal guided joint encoding feature.

[0031] It should be understood that the molten liquid state feature map is a feature representation extracted from the image of the molten liquid, including visual features such as the color, flow characteristics, and bubble distribution of the molten liquid; the melt temperature time series associated feature vector is a feature representation extracted from the temperature data in the melting process, expressing the temperature change trend over time, such as the dynamic characteristics of the heating rate and cooling rate of the melting process, reflecting the thermodynamic characteristics of the melting process. In order to be able to fuse data from different modes (images and time series) to form a comprehensive feature representation, so as to capture various information in the melting process, more comprehensively and accurately describe the overall picture of the melting process, thereby improving the comprehensive understanding ability of the model and the control of the processing process, in the technical solution of the present application, the molten liquid state feature and the melt temperature time series associated feature are combined with state-temperature prompt enhancement to obtain the melt state-temperature cross-modal guided joint encoding feature.

[0032] In detail, the local feature decomposition of the molten liquid state characteristic map along the channel dimension is first performed to obtain a set of local feature matrices of the molten liquid state. In the molten liquid state characteristic map, different channels may represent different physical characteristics, such as color, bubble distribution, etc. Through local feature decomposition, the characteristics of each channel can be analyzed independently to capture more specific details. For example, a channel may focus on the distribution of bubbles, while another channel focuses on the color changes of the molten liquid. This subdivision can improve the resolution of the features, making the model more sensitive to each feature, and thus better understanding the state of the molten liquid. The above process can be expressed as: ;in, is the characteristic diagram of the molten liquid state, , , and are the first, second, and third local feature matrices of the melt state, respectively. and The local characteristic matrix of the melt state, Represents the eigendecomposition operation on the input data.

[0033] Specifically, Figure 5 FIG. 1 is a flow chart of step S142 in the automatic control method of the fly ash melting processing system according to an embodiment of the present application. Figure 5As shown, the step S142 includes: S1421, using the melt temperature time series association feature vector as the query vector and each melt liquid state local feature matrix in the set of the melt liquid state local feature matrix as the key matrix, inputting the query vector and the key matrix into a prompt learning network based on a converter structure to obtain a set of melt liquid state-temperature local query prompt semantic encoding vectors; S1422, inputting the melt temperature time series association feature vector and each melt liquid state-temperature local query prompt semantic encoding vector in the set of the melt liquid state-temperature local query prompt semantic encoding vector into a cross-modal mask weaving network based on prompt information to obtain a melt liquid state-temperature local query prompt semantic encoding vector based on prompt information. S1423, calculating the set of molten liquid state-temperature cross-modal local feature mask weight matrices based on prompt information and the set of molten liquid state local feature matrices, and multiplying the molten liquid state-temperature cross-modal local granularity significant interaction matrices by the position point between each group of corresponding molten liquid state-temperature cross-modal local feature mask weight matrices based on prompt information and the molten liquid state local feature matrices; S1424, performing feature aggregation on the set of molten liquid state-temperature cross-modal local granularity significant interaction matrices to obtain a molten state-temperature cross-modal guided joint coding feature map as the molten state-temperature cross-modal guided joint coding feature.

[0034] More specifically, in the embodiment of the present application, the step S1421 includes: calculating the similarity score between the melt temperature time series associated feature vector and each row vector in the melt state local feature matrix to obtain a set of melt state-temperature similarity scores; normalizing each melt state-temperature similarity score in the set of melt state-temperature similarity scores to obtain a set of melt state-temperature weight coefficients; calculating the weighted sum of each melt state-temperature weight coefficient in the set of melt state-temperature weight coefficients and each row vector in the melt state local feature matrix to obtain the melt state-temperature local query hint semantic encoding vector. The above process can be expressed as: ;in, is the first in the set of local characteristic matrices of the molten liquid state The local characteristic matrix of the melt state, and They are The first, second, and and row vectors, for coding, is the melting temperature time series associated eigenvector, yes The transposed vector of is matrix multiplication, is the two-norm of the vector, for and The melt state-temperature similarity score between represents the exponential function value with the natural constant e as the base, yes Middle row vectors, for and The melt state-temperature similarity score between for The corresponding melt state-temperature weight coefficient, yes The number of row vectors in yes The corresponding melt state-temperature local query hint semantic encoding vector, Represents point product by position.

[0035] That is, the melting temperature time series associated feature vector is used as a query vector, and each matrix in the set of molten liquid state local feature matrices is used as a key matrix, and the two are sent to the prompt learning network based on the converter structure. The core of the converter architecture is to use the self-attention mechanism to enable the network to dynamically pay attention to different features in the molten liquid state local feature matrix when processing the melting temperature time series associated feature vector, and then identify the correlation between the melting temperature and the different features of the molten liquid state, dynamically select which features are the most important and adjust the weight accordingly, thereby enhancing the model's understanding of the relationship between the two different modal data of temperature and melting state, ensuring that these relationships can be effectively used in subsequent processing, thereby generating a series of molten liquid state-temperature local query prompt semantic encoding vectors containing key information of inter-modal interaction.

[0036] Then, each melt state-temperature local query prompt semantic encoding vector in the set of the melt temperature time series association feature vector and the melt state-temperature local query prompt semantic encoding vector is input into a cross-modal mask weaving network based on prompt information. The cross-modal mask weaving network can identify and highlight the important interaction between the melt temperature and the melt state. This means that in the process of processing data, we can focus on those features that have the greatest impact on the melting process, thereby improving the understanding of key temperature changes and the molten liquid state. And by generating a mask weight matrix, the network can effectively screen out key features related to the melting process, suppress irrelevant or redundant information, and ensure that the features obtained in the final output are clearer, thereby improving the robustness of the model. The above process can be expressed as: ;in, is the melting temperature time series associated eigenvector, yes The corresponding melt state-temperature local query hint semantic encoding vector, for The transposed vector of represents matrix multiplication, for and The scale of the melt state-temperature correlation matrix obtained by multiplication is the length of the matrix multiplied by the width of the matrix. yes function, yes The corresponding melt state-temperature cross-modal local feature mask weight matrix based on the prompt information.

[0037] Subsequently, the set of the molten liquid state-temperature cross-modal local feature mask weight matrices based on the prompt information and each corresponding set of the molten liquid state-temperature cross-modal local feature mask weight matrices based on the prompt information and the molten liquid state local feature matrices in the set are subjected to positional point multiplication operations to obtain a set of molten liquid state-temperature cross-modal local granularity significant interaction matrices. The point multiplication operation enables the high-weight features in the mask weight matrix to be enhanced with the molten liquid state features to form a molten liquid state-temperature cross-modal local granularity significant interaction matrix with higher information density. This high information density feature representation can better capture subtle differences and important interactions between modalities. Finally, by performing feature aggregation on the set of molten liquid state-temperature cross-modal local granularity significant interaction matrices, a molten state-temperature cross-modal guided joint encoding feature map is obtained. The above process can be expressed as: ;in, , , and are the first, second, and third local feature matrices of the melt state, respectively. and The local characteristic matrix of the melt state, , , and yes , , and The corresponding melt state-temperature cross-modal local feature mask weight matrix based on the prompt information, is the feature concatenation along the channel dimension, is the melting state-temperature cross-modal guided joint encoding feature map, Represents point product by position.

[0038] In step S150, based on the molten state-temperature cross-modal guided joint coding feature, a control result is obtained, and the control result is used to represent the recommended decoding value of the heating power at the next time point. Specifically, in an embodiment of the present application, the step S150 includes: passing the molten state-temperature cross-modal guided joint coding feature map through a decoder-based heating power controller to obtain the control result. It should be understood that in specific technical application fields, such as industrial automation control, the decoder plays a vital role, especially when it comes to heat management of complex systems such as melting processes. The main task of the decoder is to extract useful information from complex, multi-dimensional data sets, and then guide decision-making in actual operations, such as determining the most suitable heating power. This process first requires converting a high-dimensional feature map containing key information such as the state of the molten liquid and temperature changes into a one-dimensional vector. This step is intended to simplify the data structure and make subsequent processing more efficient and feasible. The converted vector is then input into a decoder, which is usually composed of multiple processing layers, which may include but are not limited to fully connected layers, recurrent neural network layers, etc. Each processing layer has a specific task, such as further extraction and abstraction of features, ensuring that deep patterns and regularities can be captured from the raw data. As the data passes through these layers, a high-level feature representation is gradually formed, which can better reflect the essential characteristics of the input data and is crucial for subsequent decision support. Finally, the decoder generates a continuous recommendation for heating power through an output layer. This output layer usually uses a linear activation function to directly output specific values ​​rather than probability distributions. The recommended value of heating power is directly related to the safety and efficiency of the melting process, so its accuracy is extremely important. To achieve this, the decoder needs to be trained based on a large amount of historical data. In this process, the model learns how to effectively extract relevant information from the feature map to minimize the difference between the predicted value and the actual required heating power. Commonly used evaluation criteria include loss functions such as mean square error, which can quantify the accuracy of the prediction and guide the optimization of model parameters. The entire training process is an iterative optimization process, which gradually improves the performance of the decoder by continuously adjusting the parameters within the model. Once trained, the decoder can play an important role in practical applications, not only improving production efficiency, but also ensuring the safety of operation, reducing energy waste, and promoting sustainable development. In short, the conversion process from the feature map to the recommended heating power value reflects the depth and breadth of the application of artificial intelligence technology in modern industrial control technology, and is one of the key steps to achieve intelligent manufacturing. In this way, by converting the melt state-temperature cross-modal guided joint encoding feature map into a specific recommended heating power value through the decoder, the heating power can be dynamically adjusted to ensure that the state and temperature of the melt are always in the optimal state.

[0039] In particular, when the molten liquid state feature map and the molten temperature time-series correlation feature vector respectively represent the image semantic features of the molten liquid state image and the time-series correlation features of the molten temperature in the melting furnace, during cross-modal joint encoding assisted by continuous prompts, the molten state-temperature cross-modal guided joint encoding feature map will also have joint encoding differences of different modal features, thereby causing distribution diversity of the joint encoding feature representation. As a result, when the molten state-temperature cross-modal guided joint encoding feature map is decoded through a decoder-based heating power controller for probability regression, it will affect the accuracy of the decoding result.

[0040] Based on this, when the melting state-temperature cross-modal guided joint coding feature map is decoded and regressed through a decoder-based heating power controller, the melting state-temperature cross-modal guided joint coding feature map is first optimized, and the optimization process specifically includes: determining the mean and variance of the feature set composed of the feature values ​​of the melting state-temperature cross-modal guided joint coding feature map, and dividing the mean by the variance to obtain the melting state-temperature cross-modal guided joint coding probability weight value: The above process can be expressed as: ;in, is the mean of the feature set consisting of the feature values ​​of the melting state-temperature cross-modal guided joint encoding feature map, is the variance of the feature set composed of the feature values ​​of the melting state-temperature cross-modal guided joint encoding feature map, A probability weight value is provided for the melting state-temperature cross-modal guided joint encoding.

[0041] The melting state-temperature cross-modal guided joint encoding feature vector obtained by expanding the melting state-temperature cross-modal guided joint encoding feature map is dot-multiplied by the inverse of the maximum eigenvalue of the melting state-temperature cross-modal guided joint encoding feature map to obtain the melting state-temperature cross-modal guided joint encoding probability constraint vector: ;in, is the melting state-temperature cross-modal guided joint encoding feature vector, represents the inverse of the maximum eigenvalue in the melting state-temperature cross-modal guided joint encoding eigenvector, It means point multiplication by position. A probability constraint vector is encoded for the melting state-temperature cross-modal guided joint.

[0042] The melting state-temperature cross-modal guided joint coding probability constraint vector is added with the melting state-temperature cross-modal guided joint coding probability weight value, and the logarithmic value with base 2 is calculated to obtain the melting state-temperature cross-modal guided joint coding information interaction vector: ;in, is the melting state-temperature cross-modal guided joint encoding probability constraint vector, is the melting state-temperature cross-modal guided joint encoding probability weight value, To add by location, is the logarithmic function with base 2, A cross-modal guided joint encoding information interaction vector is provided for the melting state-temperature.

[0043] After subtracting the melting state-temperature cross-modal guided joint coding probability constraint vector from the melting state-temperature cross-modal guided joint coding probability weight value, the inverse of each eigenvalue is calculated to obtain the melting state-temperature cross-modal guided joint coding sequence constraint vector: ;in, is the melting state-temperature cross-modal guided joint encoding probability constraint vector, To subtract by position point, is the melting state-temperature cross-modal guided joint encoding probability weight value, To calculate the inverse of each eigenvalue in a vector, A sequence constraint vector is encoded for the melting state-temperature cross-modal guided joint.

[0044] Finally, the melting state-temperature cross-modal guided joint coding information interaction vector is added with the melting state-temperature cross-modal guided joint coding sequence constraint vector to obtain an optimized melting state-temperature cross-modal guided joint coding feature vector, and the optimized melting state-temperature cross-modal guided joint coding feature vector is restored to the optimized melting state-temperature cross-modal guided joint coding feature map.

[0045] That is, considering that the weight matrix during decoding probability regression acts on the molten state-temperature cross-modal guided joint coding feature vector expanded by the molten state-temperature cross-modal guided joint coding feature map, the distribution diversity of the molten state-temperature cross-modal guided joint coding feature vector causes the weight stimulus parameter uncertainty of the weight matrix, so that the molten state-temperature cross-modal guided joint coding feature vector is missing through the posterior probability density inference of the weight matrix, which will affect the accuracy of the probability regression result. The above optimization process simulates the mesoscale interaction structure under probability constraints between the eigenvalue scale and the eigenvector scale of the molten state-temperature cross-modal guided joint coding feature vector through the probability statistical characteristics of the molten state-temperature cross-modal guided joint coding feature map, so as to construct a bidirectional hidden variable motif based on the mesoscale short sequence relative to the probability statistical value to perform mesoscale bidirectional migration, and perform posterior recovery based on the sequence constraint on the interactive information, so as to improve the convergence effect of the molten state-temperature cross-modal guided joint coding feature vector in the probability density domain, and improve the accuracy of the control result obtained by the decoder-based heating power controller. In this way, compared with the traditional fixed temperature setting value, the system can analyze the state of the melting process in real time, and dynamically and accurately adjust the heating power to ensure that the temperature always remains within the optimal range, thereby reducing temperature fluctuations, maintaining the stability of the melting process, and realizing intelligent monitoring and control of the fly ash melting process.

[0046] In summary, the automatic control method of the fly ash melting processing system based on the embodiment of the present application is explained, which uses an image analysis and data processing algorithm based on artificial intelligence to perform state analysis of the molten liquid state image and time series correlation of the melting temperature data in the melting furnace, so as to intelligently obtain the recommended decoding value of the heating power at the next time point according to the prompt enhanced joint representation between the molten liquid state characteristics and the melting temperature time series correlation characteristics. In this way, compared with the traditional fixed temperature setting value, the system can analyze the state of the melting process in real time, and dynamically and accurately adjust the heating power to ensure that the temperature is always kept within the optimal range, thereby reducing temperature fluctuations, maintaining the stability of the melting process, and realizing intelligent monitoring and control of the fly ash melting process.

[0047] The present application also provides an automatic control system for a fly ash melting processing system, which is used to execute the automatic control method for the fly ash melting processing system as described above. Figure 6 FIG. 1 is a block diagram of an automatic control system of a fly ash melting processing system according to an embodiment of the present application. Figure 6As shown, the automatic control system 100 of the fly ash melting processing system according to the embodiment of the present application includes: a data acquisition module 110, which is used to obtain a molten liquid state image collected by a high-temperature camera deployed at the observation window of the melting furnace, and obtain a time series of the melting temperature data in the melting furnace; a molten liquid state feature extraction module 120, which is used to extract the state feature of the molten liquid state image to obtain the molten liquid state feature; a molten temperature time series feature extraction module 130, which is used to extract the time series feature of the time series of the molten temperature data in the melting furnace to obtain the molten temperature time series correlation feature; a molten state-temperature cross-modal joint encoding module 140, which is used to perform state-temperature prompt enhancement on the molten liquid state feature and the molten temperature time series correlation feature to obtain a molten state-temperature cross-modal guidance joint encoding feature; a heating power recommendation result generation module 150, which is used to obtain a control result based on the molten state-temperature cross-modal guidance joint encoding feature, and the control result is used to represent the recommended decoding value of the heating power at the next time point.

[0048] Here, those skilled in the art will appreciate that the specific functions and operations of the various units and modules in the automatic control system 100 of the fly ash melting processing system have been described in detail above. Figures 3 to 5 The automatic control method of the fly ash melting processing system has been described in detail, and therefore, its repeated description will be omitted.

[0049] In summary, the automatic control system 100 of the fly ash melting processing system based on the embodiment of the present application is explained, which uses an image analysis and data processing algorithm based on artificial intelligence to perform state analysis of the molten liquid state image and time series association of the melting temperature data in the melting furnace, so as to intelligently obtain the recommended decoding value of the heating power at the next time point according to the prompt enhanced joint representation between the molten liquid state characteristics and the melting temperature time series association characteristics. In this way, compared with the traditional fixed temperature setting value, the system can analyze the state of the melting process in real time, and dynamically and accurately adjust the heating power to ensure that the temperature is always kept within the optimal range, thereby reducing temperature fluctuations, maintaining the stability of the melting process, and realizing intelligent monitoring and control of the fly ash melting process.

Claims

1. An automatic control method for a fly ash melting treatment system, the fly ash melting treatment process comprising the following steps: Washing the fly ash to be treated with water to obtain washed fly ash; Drying the washed fly ash to obtain dried washed fly ash; The dried washed fly ash is mixed with solid waste to form a shape, and then transported to a melting furnace together with fuel carbon for high-temperature melting treatment to form a molten liquid and molten waste gas; the molten liquid is water quenched to form a glass product; the glass product is screened and packaged; it is characterized in that the automatic control method of the fly ash melting treatment system is applied to the steps: the dried washed fly ash is mixed with solid waste to form a shape, and then transported to a melting furnace together with fuel carbon for high-temperature melting treatment to form a molten liquid and molten waste gas; wherein the automatic control method of the fly ash melting treatment system comprises: obtaining a high-pressure gas disposed at an observation window of the melting furnace; A temperature camera captures a molten liquid state image, and obtains a time series of molten temperature data in the molten furnace; extracts state features from the molten liquid state image to obtain molten liquid state features; extracts time series features from the time series of molten temperature data in the molten furnace to obtain molten temperature time series correlation features; performs state-temperature prompt enhancement on the molten liquid state features and the molten temperature time series correlation features to obtain molten state-temperature cross-modal guidance joint coding features; based on the molten state-temperature cross-modal guidance joint coding features, obtains a control result, and the control result is used to represent a recommended decoding value of the heating power at the next time point.

2. The automatic control method of the fly ash melting processing system according to claim 1, characterized in that: The state feature extraction is performed on the molten liquid state image to obtain the molten liquid state feature, comprising: passing the molten liquid state image through a molten liquid state feature extractor based on a depth model to obtain a molten liquid state feature map as the molten liquid state feature.

3. The automatic control method of the fly ash melting processing system according to claim 2, characterized in that: The molten liquid state feature extractor based on the deep model is a molten liquid state feature extractor based on a deep separable convolutional neural network model.

4. The automatic control method of the fly ash melting processing system according to claim 3, characterized in that: The time series of the melting temperature data in the melting furnace is subjected to time series feature extraction to obtain a melting temperature time series correlation feature, including: passing the time series of the melting temperature data in the melting furnace through a melting temperature time series feature extractor based on a forward LSTM model to obtain a melting temperature time series correlation feature vector as the melting temperature time series correlation feature.

5. The automatic control method of the fly ash melting processing system according to claim 4, characterized in that: The molten liquid state feature and the molten temperature time series associated feature are combined with state-temperature hint enhancement to obtain a molten state-temperature cross-modal guided joint coding feature, including: performing local feature decomposition along the channel dimension on the molten liquid state feature map to obtain a set of molten liquid state local feature matrices; performing hint-assisted cross-modal guided combination on the set of the molten temperature time series associated feature vector and the molten liquid state local feature matrix to obtain the molten state-temperature cross-modal guided joint coding feature.

6. The automatic control method of the fly ash melting processing system according to claim 5, characterized in that: The set of the melt temperature time series associated feature vector and the melt state local feature matrix is ​​subjected to a hint-assisted cross-modal guided joint to obtain the melt state-temperature cross-modal guided joint coding feature, including: using the melt temperature time series associated feature vector as a query vector and each melt state local feature matrix in the set of the melt state local feature matrix as a key matrix, inputting the query vector and the key matrix into a hint learning network based on a converter structure to obtain a set of melt state-temperature local query hint semantic coding vectors; inputting each melt state-temperature local query hint semantic coding vector in the set of the melt temperature time series associated feature vector and the melt state-temperature local query hint semantic coding vector into a hint-based learning network. A cross-modal mask weaving network of information is used to obtain a set of molten liquid state-temperature cross-modal local feature mask weight matrices based on the prompt information; the set of molten liquid state-temperature cross-modal local feature mask weight matrices based on the prompt information and the set of molten liquid state local feature matrices are calculated by multiplying the position points between each group of corresponding molten liquid state-temperature cross-modal local feature mask weight matrices based on the prompt information and the molten liquid state local feature matrices to obtain a set of molten liquid state-temperature cross-modal local granularity significant interaction matrices; the set of molten liquid state-temperature cross-modal local granularity significant interaction matrices is feature aggregated to obtain a molten state-temperature cross-modal guided joint coding feature map as the molten state-temperature cross-modal guided joint coding feature.

7. The automatic control method of the fly ash melting processing system according to claim 6, characterized in that: Taking the melt temperature time series associated feature vector as the query vector and taking each melt state local feature matrix in the set of melt state local feature matrices as the key matrix, the query vector and the key matrix are input into a prompt learning network based on a converter structure to obtain a set of melt state-temperature local query prompt semantic encoding vectors, including: calculating the similarity score between the melt temperature time series associated feature vector and each row vector in the melt state local feature matrix to obtain a set of melt state-temperature similarity scores; normalizing each melt state-temperature similarity score in the set of melt state-temperature similarity scores to obtain a set of melt state-temperature weight coefficients; calculating the weighted sum of each melt state-temperature weight coefficient in the set of melt state-temperature weight coefficients and each row vector in the melt state local feature matrix to obtain the melt state-temperature local query prompt semantic encoding vector.

8. The automatic control method of the fly ash melting processing system according to claim 7, characterized in that: Based on the molten state-temperature cross-modal guided joint coding feature, a control result is obtained, including: passing the molten state-temperature cross-modal guided joint coding feature map through a decoder-based heating power controller to obtain the control result.

9. An automatic control system for a fly ash melting processing system, used to execute the automatic control method for a fly ash melting processing system according to claim 1, characterized in that: The automatic control system of the fly ash melting processing system includes: a data acquisition module, which is used to obtain a molten liquid state image collected by a high-temperature camera deployed at the observation window of the melting furnace, and obtain a time series of melting temperature data in the melting furnace; a molten liquid state feature extraction module, which is used to extract state features of the molten liquid state image to obtain molten liquid state features; a molten temperature time series feature extraction module, which is used to extract time series features of the time series of molten temperature data in the melting furnace to obtain molten temperature time series correlation features; a molten state-temperature cross-modal joint encoding module, which is used to perform state-temperature prompt enhancement on the molten liquid state features and the molten temperature time series correlation features to obtain a molten state-temperature cross-modal guidance joint encoding feature; a heating power recommendation result generation module, which is used to obtain a control result based on the molten state-temperature cross-modal guidance joint encoding feature, and the control result is used to represent the recommended decoding value of the heating power at the next time point.

10. The automatic control system of the fly ash melting processing system according to claim 9, characterized in that: The molten liquid state feature extraction module is used to: pass the molten liquid state image through a molten liquid state feature extractor based on a depth model to obtain a molten liquid state feature map as the molten liquid state feature.

Images