A resin melting furnace and a temperature control method for preventing resin coking

By combining visual images and temperature sensing data into a deep learning network, high-precision reconstruction of the global temperature field and prediction of heat migration during the rosin heating process are achieved, solving the problem of coking during rosin heating and realizing uniform heating and efficient melting in the furnace.

CN123114110APending Publication Date: 2026-11-17HUNAN SHUNXIN METAL PRODUCTS TECHNOLOGY CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611332625.7
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-31
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing methods for controlling the temperature of rosin heating cannot achieve full-domain temperature field perception, solid-liquid phase identification, thermal migration prediction based on viscosity-temperature characteristics, and differentiated regional control based on global energy conservation. This results in rosin easily coking during heating, affecting production efficiency and safety.

Method used

By fusing visual images and temperature sensing data, and combining the unique physicochemical properties of pine resin, a deep learning network structure is used to reconstruct the global temperature field and predict heat migration trends. Differentiated regional control based on global energy conservation is implemented to prevent local coking.

Benefits of technology

It achieves high-precision reconstruction of the entire temperature field and advanced and accurate prediction of heat transfer during the rosin heating process, ensuring uniform heating in the furnace, completely eliminating local coking, and taking into account efficient melting, thereby improving production efficiency and safety.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123114110A_ABST
    Figure CN123114110A_ABST
Patent Text Reader

Abstract

This invention discloses a rosin melting furnace and a temperature control method for preventing rosin coking, comprising: using phase change temperature as a physical constraint boundary to intervene in the network output, avoiding outliers driven by pure data, and reconstructing a temperature field that conforms to thermodynamic laws; converting temperature into heat transfer modulation weights and injecting them into the prediction network to adaptively simulate heat diffusion under viscosity-temperature changes, giving the model physical foresight to accurately predict heat migration; based on the above support, reducing power in the risk zone and performing energy conservation compensation in the low-temperature zone. Each link is closely linked, progressing from perception and prediction to control, achieving comprehensive control that prevents coking and achieves efficient melting.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the technical field of rosin dissolution, and more particularly to a rosin dissolution furnace and a temperature control method for preventing rosin from coking. Background Technology

[0002] Rosin, as an important natural renewable resin resource, has extremely wide applications in chemical, pharmaceutical, light industry, and new materials fields. Melting and heating are indispensable basic steps in the deep processing and purification of rosin. However, rosin has unique physicochemical properties; it is solid at room temperature, gradually softens and eventually transforms into a liquid as the temperature rises. During this complex solid-liquid phase transition, rosin exhibits poor thermal conductivity and its viscosity is extremely sensitive to temperature, exhibiting a significant "viscosity-temperature characteristic." When the local temperature is too high or the heating time is too long, rosin is prone to thermal degradation and oxidative polymerization reactions, resulting in severe "coking." Coking not only severely damages the color and physicochemical quality of rosin products but also causes a hard carbonized layer to adhere to the inner wall of the reaction vessel, significantly reducing the heat transfer efficiency of the equipment and even leading to pipe blockage and serious safety accidents. Therefore, achieving precise temperature control during the rosin heating process to effectively prevent coking is a core technical challenge that the industry has long needed to address.

[0003] Currently, most common methods for controlling the heating temperature of rosin in the industry rely on traditional single-point or few-point contact temperature measurement techniques, such as installing only a few temperature sensors on the top (upper head) or side wall of the reactor. This conventional method has significant limitations: First, due to the complex solid-liquid mixture during the melting process of rosin, the temperature distribution inside the reactor exhibits strong non-uniformity and spatial nonlinearity. The readings from a few temperature sensors cannot accurately and comprehensively reflect the three-dimensional temperature field distribution throughout the furnace. Second, existing technologies lack effective visual perception of the spatial information of the rosin solid-liquid interface, making it impossible to accurately define the physical boundary between the solid and liquid phases. This results in a "blind" heating process, easily leading to continuous overheating of locally liquefied areas before the solid rosin has completely melted.

[0004] Furthermore, existing temperature control systems suffer from serious deficiencies in predicting heat migration. Traditional control logic is mostly based on simple PID feedback control, completely neglecting the unique viscosity-temperature characteristics of rosin. In reality, the local fluidity of rosin plays a decisive modulating role in the intensity of its lateral heat diffusion: in the solid phase or low-temperature, high-viscosity regions, heat is difficult to diffuse, easily leading to heat accumulation; while in high-temperature, low-viscosity regions, heat diffusion is significantly enhanced. Existing methods ignore this dynamic heat transfer difference based on phase and viscosity, failing to predict the lateral heat migration trend in complex phases, often only reacting passively after abnormally high temperatures have already occurred locally. This severe lag makes it impossible for the system to provide early warning of coking risk.

[0005] In terms of heating power control strategies, existing technologies also have significant shortcomings. When localized overheating is detected, traditional methods typically employ a crude approach of "one-size-fits-all" global cooling or direct cutting off of the heat source. While this approach can alleviate localized overheating to some extent, it leads to a sharp decrease in the overall heat input within the furnace, severely slowing down the overall melting process of the resin and significantly reducing production efficiency. Furthermore, existing methods lack refined, zone-specific control capabilities, fail to establish a dynamic power allocation mechanism based on coking risk classification, and are unable to provide reasonable power compensation to low-temperature safe zones while reducing heating power in high-risk areas. This results in extremely uneven heating within the furnace, failing to meet the process requirements of maintaining overall heat balance and efficient melting while ensuring a safety baseline against coking.

[0006] In summary, existing methods for controlling the heating temperature of rosin have significant technological gaps in areas such as global temperature field sensing, solid-liquid phase identification, thermal migration prediction based on viscosity-temperature characteristics, and differentiated regional control based on global energy conservation, making it difficult to meet the needs of modern high-quality rosin processing. Summary of the Invention

[0007] In view of this, the present invention provides a rosin melting furnace and a temperature control method for preventing rosin coking. The purpose is to improve the deep learning network structure by fusing visual images and temperature sensing data and combining the unique physicochemical properties of rosin, so as to achieve high-precision reconstruction of the temperature field of the whole furnace and accurate prediction of heat migration trend. Then, based on the prediction results, differentiated zone control with global energy conservation is implemented, and finally, a comprehensive control goal of uniform heating in the furnace, complete elimination of local coking and efficient melting is achieved.

[0008] To achieve the above objectives, this invention discloses a rosin melting furnace, comprising a support (1), a lower end cap (2), a cylinder (3), and an upper end cap (5). The lower end cap (2) is fixedly installed on the top of the support (1). The bottom of the cylinder (3) is sealed to the top of the lower end cap (2). The top of the cylinder (3) and the bottom of the upper end cap (5) are sealed together to form a closed melting chamber. The top surface of the upper end cap (5) is provided with a flange connector 2 (9), a flange connector 3 (10), a lifting lug (4), and a pressure gauge (11). The side of the upper end cap (5) The cylinder (3) is provided with a vent hole (13). The right side wall of the cylinder (3) is provided with a nameplate (6) and a sight glass (12) from bottom to top. The left side wall of the lower head (2) is provided with a flange pipe (8). The side wall of the support (1) is fixedly provided with a grounding plate (7). The top end face of the upper head (5) is reserved with multiple sets of functional interfaces, including a discharge port (14), a spare port one (15), a spare port two (16), a safety valve port (17), a pressure gauge port (18), a vent (19), a sight glass port (20), and a thermometer connector (21).

[0009] Preferably, the lifting lugs (4) are symmetrically welded and fixed to the top two sides of the upper end cap (5), and the base plate (7) is fixedly installed on the side of the support (1).

[0010] Preferably, the sight glass (12) is sealed and fitted to the upper side wall of the cylinder (3), and the sight glass (12) corresponds to the sight glass opening (20) at the top of the upper end cap (5).

[0011] Preferably, the pressure gauge (11) is installed inside the pressure gauge port (18) of the upper end cap (5), and the safety valve is installed inside the safety valve port (17).

[0012] Preferably, a temperature sensor is installed inside the temperature gauge connector (21), and the nameplate (6) is fixed to the right side wall of the cylinder (3).

[0013] The present invention also provides a temperature control method for preventing resin from scorching, comprising the following steps: S1: Extract the temperature reading of the temperature sensor at the top of the upper head at the current moment, perform solid-liquid interface detection on the furnace image taken by the sight glass installed on the side wall of the cylinder, take the solid-liquid phase transition temperature of rosin as the physical constraint boundary, take the temperature reading as the global reference anchor point, combine the solid-liquid interface spatial information of the sight glass image to reconstruct the global temperature field, and generate the global high-resolution temperature field of the furnace at the current moment. S2: Will continue The high-resolution temperature field across the entire furnace is arranged chronologically to form a temporal input. Based on the viscosity-temperature characteristics of rosin, the current temperature of each spatial node is converted into a heat transfer modulation weight. This heat transfer modulation weight is then injected into the lateral heat diffusion and propagation process of the viscosity-temperature modulated heat migration prediction network, outputting the future... Predicted sequence of global temperature field within the frame furnace; S3: Extract the temperature peak and continuous temperature rise rate of each spatial node in the predicted temperature field sequence. According to the double-layer coking risk judgment criterion, classify and mark each spatial node as a high-risk coking node or a coking early warning node. Aggregate all marked nodes into coking risk areas according to spatial connectivity. Map the spatial coordinates of each coking risk area to the corresponding heating zone number of the furnace body and output a graded coking risk heating zone list. S4: Implement differentiated heat source input power regulation for each heating zone based on the coking risk level. For low-temperature heating zones not included in the coking risk list, proportionally compensate and increase the heat source input power to meet the overall heat conservation constraint in the furnace and achieve uniform heating and anti-coking temperature control in the furnace.

[0014] As a further improvement of the present invention: Optionally, step S1 further includes: Extract the temperature sensor installed inside the temperature gauge connector on the top of the upper end cap at the current moment. The temperature reading is recorded as the global reference temperature. Simultaneously, an image acquisition device installed at the sight glass on the side wall of the cylinder acquires images at any time. The captured images inside the furnace are input into a phase boundary recognition network composed of a convolutional feature extraction layer and a boundary contour regression layer, and the solid-liquid interface contour is output. Based on the pre-defined projection mapping relationship between the sight glass image coordinate system and the furnace three-dimensional space coordinate system, the solid-liquid interface contour is transformed from the image coordinate system to the furnace three-dimensional space coordinate system, thereby determining the set of spatial nodes covered by the visible area of ​​the sight glass in the furnace three-dimensional space. ;right Within each spatial node, nodes within the area enclosed by the solid-liquid interface contour are labeled as resin solid phase nodes, and nodes outside the contour are labeled as resin liquid phase nodes; for nodes not belonging to... The furnace interior space nodes, based on the global reference temperature Solid-liquid phase transition temperature of rosin Phase determination is performed based on the magnitude relationship: when Below When all non-visible spatial nodes are labeled as resin solid phase nodes; when Not less than At that time, all non-visible spatial nodes were labeled as resin liquid phase nodes; all The set of all solid-phase nodes in a given spatial node is denoted as the resin solid-phase region set. Let the set of all liquid phase nodes be denoted as the set of rosin liquid phase regions. ; Based on the solid-liquid phase transition temperature of rosin As a physical constraint boundary, for the entire furnace area Phase constraints are imposed on the reconstruction temperature of each spatial node: solid-phase region set The upper limit of the reconstruction temperature of each spatial node does not exceed Liquid phase region collection The lower bound of the reconstruction temperature of each spatial node is not lower than Under the condition that the phase state constraint is satisfied, the global reference temperature is... With the entire furnace area The phase state labeling vectors of each spatial node are jointly input into a phase state-guided denser network. This network consists of three layers of spatial convolution and bilinear interpolation upsampling, which are stacked alternately to extend the temperature readings and spatial distribution information of the phase state to the entire furnace area. Each spatial node; the phase constraint is directly applied to the output of each layer of the phase-guided denser network by a truncation operation, so that the reconstruction temperature of each node satisfies: ; in, Indicates time No. The reconstruction temperature of each spatial node. , This indicates that the phase-guided denser network is for the first... Each spatial node at time The output is the unconstrained temperature prediction. This indicates the solid-liquid phase transition temperature of rosin. Take the smaller value for the two parameters. Take the larger value of the two parameters; Reconstruction temperature of each spatial node Overall composition moment High-resolution temperature field throughout the furnace .

[0015] Optionally, step S2 further includes: Time Previous consecutive Full-range high-resolution temperature field within the frame furnace Arranged in chronological order, they form a temporal input tensor, where To preset the number of historical frames, the temporal input tensor is input into the viscosity-temperature modulation thermal migration prediction network, which is composed of alternating stacks of spatial convolution modules that perform local receptive field convolution operations along the spatial dimension and temporal recursive modules that perform gated state updates along the temporal dimension. Based on the viscosity-temperature characteristic that the viscosity of rosin decreases monotonically with increasing local temperature, each spatial node... At any moment Reconstruction temperature Convert to heat transfer modulation weight The heat transfer modulation weight quantifies the modulation effect of the local flowability of rosin on the intensity of lateral heat diffusion, and the calculation formula is as follows: ; in, Indicates the first Each spatial node at time Heat transfer modulation weight, Indicates the first Each spatial node at time The reconstruction temperature This indicates the solid-liquid phase transition temperature of rosin. Indicates the upper limit of the furnace temperature. This is an indicator function that takes the value 1 when the condition inside the parentheses is true and 0 when it is false. The range of values ​​is When the node is in the solid phase region Lateral heat diffusion is suppressed; when the node temperature is higher than hour, It increases with increasing temperature, and its ability to diffuse heat laterally is enhanced. The heat transfer modulation weight As the channel modulation coefficients of each convolution kernel in the spatial convolution module of the viscosity-temperature modulated heat migration prediction network, they are multiplied element-wise with the current hidden layer feature vector of each spatial node and then participate in the aggregation of convolutional information between adjacent nodes, enabling the network to accurately predict lateral heat migration. Apply strong diffusion weights to value nodes and apply them to low-value nodes. A weak diffusion weight is applied to the value nodes to automatically distinguish between heat accumulation trend nodes and heat diffusion trend nodes, and the future value is output. Prediction sequence of global temperature field in frame furnace ,in This is the preset number of prediction frames.

[0016] Optionally, step S3 further includes: For the future Prediction sequence of global temperature field in frame furnace Each spatial node Extracting peak temperature With sustained rate of temperature rise Two characteristics: the peak temperature For the first Each spatial node The maximum temperature value within the frame prediction sequence; the sustained rate of temperature rise. For the first Each spatial node The slope of the linear fit of temperature change over time within the frame prediction sequence. A positive slope indicates a continuous temperature rise trend, while a zero or negative slope indicates a stable or falling temperature trend. Based on the two-layer coking risk assessment criteria, each spatial node is classified and labeled: the first layer is determined by the critical temperature of resin coking. The threshold for determining the peak temperature. Spatial nodes are directly marked as high-risk nodes for coking; the second layer of judgment is based on the warning temperature threshold. The threshold for determining the peak temperature and the threshold for the early warning temperature rise rate. The threshold for rate determination is denoted as follows: , and The spatial nodes are marked as coking warning nodes; the remaining spatial nodes are marked as normal nodes. All high-risk coking nodes and coking early warning nodes are aggregated into a set of high-risk coking areas based on three-dimensional spatial connectivity. and coking warning area set ;Will The three-dimensional spatial coordinates of nodes within each region are matched with a pre-established heating zone spatial index table. The covered heating zone number is written into the high-risk heating zone list. ;Will The covered heating zone number is written into the warning heating zone list. Output a list of heating zones with graded coking risk. .

[0017] Optionally, step S4 further includes: List of high-risk heating zones In each heating zone, power is reduced according to the proportion of high-risk zones. The input power of the partition heat source is changed by Reduce to ; List of heating zones for early warning Each heating zone in the middle will reduce power according to the warning ratio. The input power of the partition heat source is changed by Reduce to ,in Indicates the first Current heat source input power for each partition After implementing the power reduction regulation, verify whether the heat source input power after regulation in each zone is lower than the minimum power limit required to maintain the continuous dissolution of pine resin. For power levels lower than 1 after regulation The partition limits the heat source input power of the partition to The actual power reduction amount in each risk zone Defined as the current heat source input power of the partition. The difference between the final heat source input power after clamping and the actual power input after clamping, i.e. ,in The power reduction ratio corresponding to the partition is taken as follows: high-risk heating partition is... Early warning heating zone ; For heating zones not included in the graded coking risk list For the low-temperature heating zones, compensation power increases are calculated and applied based on the overall heat conservation constraint within the furnace. The conservation constraint is based on the actual power reduction of each risk zone, and its expression is as follows: ; in, Indicates the first Compensation power increase for each low-temperature heating zone; compensation power increase for each low-temperature heating zone. The heat source input power of each zone is allocated according to the proportion of its current input power to the total input power of all low-temperature heating zones, and the adjusted input power of each low-temperature heating zone must not exceed the rated power limit of that zone. When the proportionally allocated compensation power causes the input power of the heat source in a certain low-temperature zone to exceed... At that time, the partition compensation power boost is truncated to The difference between the current heat source input power of the partition and the remaining compensation power generated by the cut-off is proportionally redistributed to the remaining low-temperature heating partitions whose rated power has not yet been saturated. The equation ensures that the total total heat input in the furnace remains unchanged before and after the adjustment.

[0018] Compared with the prior art, the present invention has at least the following beneficial effects: Firstly, in terms of global temperature field reconstruction, this invention does not directly apply conventional image interpolation or standard convolutional neural networks, but innovatively constructs a phase-guided denser network. This network uses the solid-liquid phase transition temperature of rosin as a rigid physical constraint boundary, and directly intervenes in the output of each layer of the network through truncation and clipping operations. This algorithmic improvement effectively avoids the defect of pure data-driven models that are prone to outputting outliers that violate thermodynamic principles in complex industrial scenarios, so that the reconstructed high-resolution temperature field not only has extremely high spatial accuracy, but also strictly conforms to the physical objective laws of rosin melting.

[0019] Secondly, in the heat migration prediction stage, this invention substantially modifies the underlying information transmission mechanism of the spatiotemporal prediction network, taking into account the unique viscosity-temperature characteristics of rosin. This invention dynamically converts local temperature into heat transfer modulation weights and directly injects them as channel modulation coefficients of each convolution kernel in the spatial convolution module into the calculation process of lateral heat diffusion. This profound algorithmic structural improvement enables the network to adaptively simulate the real physical phenomena of heat accumulation in high-viscosity regions and easy heat diffusion in low-viscosity regions when extrapolating heat migration. This breaks the limitation of traditional time-series prediction networks that rely solely on historical data statistical patterns, giving the model true physical interpretability and foresight, thus enabling it to extremely keenly capture potential coking risk trends.

[0020] More importantly, the various innovative aspects of this invention are closely interconnected, forming a highly efficient and collaborative temperature control closed loop. The high-precision temperature field reconstruction based on phase-state physical constraints at the front end provides an absolutely reliable initial state input for subsequent thermal migration prediction. Meanwhile, the prediction network incorporating a viscosity-temperature modulation mechanism can accurately output the evolution of heat distribution over a future period, providing solid data support for subsequent assessment of double-layer coking risk. Finally, risk region mapping based on spatial connectivity aggregation directly guides the differentiated power allocation and global heat conservation compensation at the bottom layer. This deep linkage from physical constraint perception and characteristic modulation prediction to conservation compensation control enables the entire solution to both accurately extinguish localized coking risks and maintain overall heating efficiency through energy transfer compensation. The various technical features are interdependent and mutually reinforcing, fundamentally solving the industry challenge of balancing coking prevention and high energy efficiency in rosin heating. Attached Figure Description

[0021] Figure 1 This is a schematic diagram of the overall structure of the rosin melting furnace provided in Embodiment 1 of the present invention; Figure 2 This is a top view of the interface of the top end cap of the rosin melting furnace provided in Embodiment 1 of the present invention; Figure 3 This is a schematic flowchart of a temperature control method for preventing resin coking according to an embodiment of the present invention. Detailed Implementation

[0022] The present invention will be further described below with reference to the accompanying drawings, but this is not intended to limit the present invention in any way. Any modifications or substitutions made based on the teachings of the present invention shall fall within the protection scope of the present invention.

[0023] Example 1: This example provides a rosin melting furnace, such as... Figure 1As shown, it includes: a support 1, a lower end cap 2, a cylinder 3, and an upper end cap 5. The lower end cap 2 is fixedly installed on the top of the support 1. The bottom of the cylinder 3 is sealed to the top of the lower end cap 2. The top of the cylinder 3 and the bottom of the upper end cap 5 are sealed to form a closed melting cavity. The top surface of the upper end cap 5 is provided with flange pipe 2 9, flange pipe 3 10, lifting lug 4, and pressure gauge 11.

[0024] Among them, the lifting lugs 4 are symmetrically welded and fixed to the top two sides of the upper end cap 5, and the ground plate 7 is fixedly installed on the side end of the support 1.

[0025] according to Figure 2 As shown, the upper end cap 5 has a vent hole 13 on its side, the right side wall of the cylinder 3 has a nameplate 6 and a sight glass 12 arranged from bottom to top, the left side wall of the lower end cap 2 has a flange pipe 8, and the side wall of the support 1 is fixedly provided with a grounding plate 7.

[0026] according to Figure 2 As shown, the top end face of the upper end cap 5 has multiple sets of functional interfaces reserved, including discharge port 14, spare port one 15, spare port two 16, safety valve port 17, pressure gauge port 18, vent port 19, sight glass port 20, and temperature gauge connector 21.

[0027] The sight glass 12 is sealed and assembled on the upper side wall of the cylinder 3. The sight glass 12 corresponds to the sight glass opening 20 on the top of the upper head 5. The temperature sensor is installed inside the temperature gauge connector 21. The nameplate 6 is fixed to the right side wall of the cylinder 3.

[0028] The grounding plate 7, pressure gauge 11, sight glass 12, safety valve, and thermometer are all existing technologies and are references to existing technologies. Their working principles, dimensions, and models are irrelevant to the function of this application, so they will not be described in detail. The control method of this invention is through a controller. The control circuit of the controller can be implemented by those skilled in the art through simple programming. The power supply is also common knowledge in the art. Furthermore, this invention is mainly used to protect mechanical devices, so the control method and circuit connection will not be explained in detail.

[0029] Working principle: During production, solid rosin is fed into the melting chamber of the cylinder through the vent hole and feed port of the upper head 5. High-temperature heat transfer oil or steam is introduced through flange pipe 8. The rosin inside is continuously heated through the lower head 2 and the wall of the cylinder 3. The solid rosin melts into liquid rosin. The operator observes the liquid level and melting state of the rosin inside through the sight glass 12 on the side wall of the cylinder 3. The temperature gauge connector 21 provides real-time feedback on the temperature of the internal melt. The pressure gauge 11 monitors the pressure inside the furnace chamber in real-time. When the pressure exceeds the standard, the safety valve automatically releases pressure to ensure the pressure safety of the equipment. After the rosin is completely melted and meets the standard, the liquid rosin is discharged from the top outlet 14 to enter the next distillation process. Equipment cleaning, sewage discharge, and additive addition can be achieved by expanding the external pipeline through two sets of spare ports. The equipment is transported by lifting lugs 4. Static electricity is released by the grounding plate 7 throughout the production process to avoid fire safety hazards.

[0030] Example 2: A temperature control method for preventing resin scorching, such as... Figure 3 As shown, it includes the following steps: S1: Extract the temperature reading from the temperature sensor at the top of the upper end cap at the current moment. Perform solid-liquid interface detection on the furnace image captured by the sight glass installed on the side wall of the cylinder. Using the resin solid-liquid phase transition temperature as the physical constraint boundary and the temperature reading as the global reference anchor point, combine the solid-liquid interface spatial information from the sight glass image to reconstruct the global temperature field, generating the current high-resolution global temperature field inside the furnace, including: Establish a data acquisition network for temperature sensors and image acquisition devices, connect the temperature sensor installed in the temperature gauge connector at the top of the upper head with the image acquisition device installed at the sight glass on the side wall of the cylinder, and set the data sampling frequency to synchronously acquire the real-time temperature reading and the image inside the furnace. Extract the temperature sensor installed inside the temperature gauge connector on the top of the upper end cap at the current moment. The temperature reading is recorded as the global reference temperature. Simultaneously, an image acquisition device installed at the sight glass on the side wall of the cylinder acquires images at any time. The captured images inside the furnace are input into a phase boundary recognition network consisting of a convolutional feature extraction layer and a boundary contour regression layer, which outputs the solid-liquid interface contour. In this embodiment, the convolutional feature extraction layer of the phase boundary recognition network is composed of four sets of convolutional blocks stacked sequentially. Each set of convolutional blocks contains a 3×3 convolutional layer, a batch normalization layer, and a ReLU activation function layer. The number of output feature channels for the four sets of convolutional blocks is set to 32, 64, 128, and 256 respectively. The boundary contour regression layer is composed of two layers of bilinear interpolation upsampling layers and two layers of 1×1 convolutional layers stacked alternately, which gradually upsamples the feature map resolution to the original furnace image resolution. Finally, a pixel-level binary segmentation map of the solid-liquid interface contour is output through a Sigmoid activation function. The phase boundary recognition network uses manually labeled solid-liquid interface contour samples from furnace images as training data and a binary cross-entropy loss function as the training objective. In this embodiment, the Adam optimizer is used for parameter updates during training, and the learning rate is set to 1×10. -4 Batch size is set to 16, and training rounds are set to 100. Based on the pre-defined projection mapping relationship between the sight glass image coordinate system and the furnace three-dimensional space coordinate system, the solid-liquid interface contour is transformed from the image coordinate system to the furnace three-dimensional space coordinate system, thereby determining the set of spatial nodes covered by the visible area of ​​the sight glass in the furnace three-dimensional space. ;right Within each spatial node, nodes within the area enclosed by the solid-liquid interface contour are labeled as resin solid phase nodes, and nodes outside the contour are labeled as resin liquid phase nodes; for nodes not belonging to... The furnace interior space nodes, based on the global reference temperature Solid-liquid phase transition temperature of rosin Phase determination is performed based on the magnitude relationship: when Below When all non-visible spatial nodes are labeled as resin solid phase nodes; when Not less than At that time, all non-visible spatial nodes were labeled as resin liquid phase nodes; all The set of all solid-phase nodes in a given spatial node is denoted as the resin solid-phase region set. Let the set of all liquid phase nodes be denoted as the set of rosin liquid phase regions. In this embodiment, the entire furnace interior... Each spatial node adopts a three-dimensional structured mesh node that is uniformly divided along the axial, radial, and circumferential directions of the furnace body; Based on the solid-liquid phase transition temperature of rosin As a physical constraint boundary, for the entire furnace area Phase constraints are imposed on the reconstruction temperature of each spatial node: solid-phase region set The upper limit of the reconstruction temperature of each spatial node does not exceed Liquid phase region collection The lower bound of the reconstruction temperature of each spatial node is not lower than Under the condition that the phase state constraint is satisfied, the global reference temperature is... With the entire furnace area The phase state labeling vectors of each spatial node are jointly input into a phase state-guided denser network. This network consists of three layers of spatial convolution and bilinear interpolation upsampling, which are stacked alternately to extend the temperature readings and spatial distribution information of the phase state to the entire furnace area. Each spatial node; the phase constraint is directly applied to the output of each layer of the phase-guided denser network by a truncation operation, so that the reconstruction temperature of each node satisfies: ; in, Indicates time No. The reconstruction temperature of each spatial node. , This indicates that the phase-guided denser network is for the first... Each spatial node at time The output is the unconstrained temperature prediction. This indicates the solid-liquid phase transition temperature of rosin. Take the smaller value for the two parameters. Take the larger value of the two parameters; Reconstruction temperature of each spatial node Overall composition moment High-resolution temperature field throughout the furnace In this embodiment, the three spatial convolutional kernels of the phase-guided denser network are all set to 3×3×3, and the number of output feature channels are set to 64, 128, and 64 respectively. Each convolutional layer is followed by a bilinear interpolation upsampling layer with an upsampling factor of 2. The phase-guided denser network is trained using furnace-wide temperature field data generated offline from a three-dimensional heat transfer numerical simulation model (CFD) of a resin melting furnace. The simulation dataset covers three scenarios: normal steady-state melting conditions, local rapid heating conditions, and high-temperature conditions close to the coking critical temperature, to ensure that the training sample distribution matches the actual application scenario. The reference temperature true value of each spatial node is used. The temperature field results are taken from the CFD simulation model output under the corresponding operating conditions; the mean square error loss term is used. Phase constraint penalty term weighted combination For training objectives: ; in, This represents the total training loss of the phase-guided dense network. Represents the phase constraint penalty weight coefficient; mean square error loss term. The calculation formula is: ; in, Indicates time No. The true reference temperature of each spatial node; phase constraint penalty term. The calculation formula is: ; in, For indicator functions, when the first Each spatial node belongs to the set of solid regions. The value is 1 if the condition is met, and 0 otherwise. For indicator functions, when the first Each spatial node belongs to the set of liquid phase regions. The value is 1 if the condition is met, and 0 otherwise; in this embodiment, Set the value to 0.5, use the Adam optimizer to update the parameters, and set the learning rate to 5×10. -4 The batch size is set to 32, and the number of training rounds is set to 200. Optionally, to improve the robustness of the phase boundary recognition network in solid-liquid interface detection under rosin surface disturbance conditions, a surface disturbance data augmentation mechanism is introduced during the training phase: for each furnace in-training image in the training set, the data is augmented from a uniformly distributed... Pixel offsets of random sampling Random perturbations are applied to the solid-liquid interface contour annotation along the radial direction of the furnace body to generate augmented samples simulating liquid surface fluctuations; in this embodiment, Set to 10 pixels.

[0031] S2: Will continue The high-resolution temperature field across the entire furnace is arranged chronologically to form a temporal input. Based on the viscosity-temperature characteristics of rosin, the current temperature of each spatial node is converted into a heat transfer modulation weight. This heat transfer modulation weight is then injected into the lateral heat diffusion and propagation process of the viscosity-temperature modulated heat migration prediction network, outputting the future... The predicted sequence of the global temperature field within the frame furnace includes: Time Previous consecutive Full-range high-resolution temperature field within the frame furnace Arranged in chronological order, they form a temporal input tensor, where To preset the number of historical frames, the temporal input tensor is input into a viscosity-temperature modulation thermal migration prediction network. This network is composed of alternating stacks of spatial convolution modules that perform local receptive field convolution operations along the spatial dimension and temporal recursive modules that perform gated state updates along the temporal dimension. In this embodiment, The viscosity-temperature modulation thermal migration prediction network is composed of three spatial convolutional modules and two temporal recursive modules stacked alternately. The number of output feature channels of the three spatial convolutional modules is set to 64, 128 and 64 respectively, and the kernel size is set to 3×3×3. The two temporal recursive modules both use gated recurrent units, and the hidden layer dimension is set to 128. Based on the viscosity-temperature characteristic that the viscosity of rosin decreases monotonically with increasing local temperature, each spatial node... At any moment Reconstruction temperature Convert to heat transfer modulation weight The heat transfer modulation weight quantifies the modulation effect of the local flowability of rosin on the intensity of lateral heat diffusion, and the calculation formula is as follows: ; in, Indicates the first Each spatial node at time Heat transfer modulation weight, Indicates the first Each spatial node at time The reconstruction temperature This indicates the solid-liquid phase transition temperature of rosin. Indicates the upper limit of the furnace temperature. This is an indicator function that takes the value 1 when the condition inside the parentheses is true and 0 when it is false. The range of values ​​is When the node is in the solid phase region Lateral heat diffusion is suppressed; when the node temperature is higher than hour, It increases with increasing temperature, and its ability to diffuse heat laterally is enhanced. The heat transfer modulation weight As the channel modulation coefficients of each convolution kernel in the spatial convolution module of the viscosity-temperature modulated heat migration prediction network, they are multiplied element-wise with the current hidden layer feature vector of each spatial node and then participate in the aggregation of convolutional information between adjacent nodes, enabling the network to accurately predict lateral heat migration. Apply strong diffusion weights to value nodes and apply them to low-value nodes. A weak diffusion weight is applied to the value nodes to automatically distinguish between heat accumulation trend nodes and heat diffusion trend nodes, and the future value is output. Prediction sequence of global temperature field in frame furnace ,in The preset prediction frame number; in this embodiment, The value is set to 5; the training of the viscosity-temperature modulation thermal migration prediction network uses high-resolution time-series temperature field data of the entire furnace domain generated by CFD offline calculation. The simulation dataset has the same operating condition coverage as described in step S1, including normal steady-state dissolution conditions, local rapid heating conditions, and high-temperature conditions close to the coking critical temperature, to ensure that the training data distribution matches the actual application scenario of coking risk prediction; the true temperature values ​​of each spatial node. The temperature field results at the corresponding time point are taken from the output of the CFD simulation model; the mean square error between the predicted frame temperature field and the actual temperature field is used as the training loss function; in this embodiment, the Adam optimizer is used for parameter updates, and the learning rate is set to 1×10. -3 The batch size was set to 16, the training rounds were set to 300, and a cosine annealing learning rate scheduling strategy was used to dynamically adjust the learning rate during training.

[0032] Optionally, to further improve the prediction accuracy of the viscosity-temperature modulated thermal migration prediction network under rapid heating conditions, a temperature rise gradient-aware weighting term is introduced into the training loss function. During training, spatial nodes with higher temperature rise rates are given greater loss weights; the total training loss after introducing the weighting term... for: ; in, This represents the total training loss after incorporating temperature rise gradient-aware weighting. Represents the temperature rise gradient-sensing weighted coefficients; temperature rise gradient-sensing weighted loss term The calculation formula is: ; in, Indicates the first Each spatial node in the predicted frame sequence number The temperature gradient sensing weight at a given location is defined as follows: ; in, This represents the scaling factor for the temperature rise gradient weights. Indicates the sampling time interval between adjacent frames. Indicates the first Each spatial node at time The actual temperature value; in this embodiment, Set to 0.2, Set to 0.05.

[0033] S3: Extract the peak temperature and sustained temperature rise rate from each spatial node in the predicted temperature field sequence. Based on the double-layer coking risk judgment criterion, classify and mark each spatial node as a high-risk coking node or a coking warning node. Aggregate all marked nodes into coking risk regions according to spatial connectivity. Map the spatial coordinates of each coking risk region to the corresponding heating zone number of the furnace body, and output a graded coking risk heating zone list, including: For the future Prediction sequence of global temperature field in frame furnace Each spatial node Extracting peak temperature With sustained rate of temperature rise Two characteristics: the peak temperature For the first Each spatial node The maximum temperature value within the frame prediction sequence; the sustained rate of temperature rise. For the first Each spatial node The slope of the linear fit of temperature change over time within the frame prediction sequence. A positive slope indicates a continuous temperature rise trend, while a zero or negative slope indicates a stable or falling temperature trend. Based on the two-layer coking risk assessment criteria, each spatial node is classified and labeled: the first layer is determined by the critical temperature of resin coking. The threshold for determining the peak temperature. Spatial nodes are directly marked as high-risk nodes for coking; the second layer of judgment is based on the warning temperature threshold. The threshold for determining the peak temperature and the threshold for the early warning temperature rise rate. The threshold for rate determination is denoted as follows: , and The spatial nodes are marked as coking warning nodes; the remaining spatial nodes are marked as normal nodes. All high-risk coking nodes and coking early warning nodes are aggregated into a set of high-risk coking areas based on three-dimensional spatial connectivity. and coking warning area set ;Will The three-dimensional spatial coordinates of nodes within each region are matched with a pre-established heating zone spatial index table. The covered heating zone number is written into the high-risk heating zone list. ;Will The covered heating zone number is written into the warning heating zone list. Output a list of heating zones with graded coking risk. .

[0034] S4: Implement differentiated heat source input power regulation for each heating zone based on the coking risk level. For low-temperature heating zones not included in the coking risk list, proportionally increase the heat source input power to meet the overall heat conservation constraint within the furnace, achieving uniform heating and anti-coking temperature control within the furnace, including: List of high-risk heating zones In each heating zone, power is reduced according to the proportion of high-risk zones. The input power of the partition heat source is changed by Reduce to ; List of heating zones for early warning Each heating zone in the middle will reduce power according to the warning ratio. The input power of the partition heat source is changed by Reduce to ,in Indicates the first Current heat source input power for each partition After implementing the power reduction regulation, verify whether the heat source input power after regulation in each zone is lower than the minimum power limit required to maintain the continuous dissolution of pine resin. For power levels lower than 1 after regulation The partition limits the heat source input power of the partition to The actual power reduction amount in each risk zone Defined as the current heat source input power of the partition. The difference between the final heat source input power after clamping and the actual power input after clamping, i.e. ,in The power reduction ratio corresponding to the partition is taken as follows: high-risk heating partition is... Early warning heating zone In this embodiment, Set to 0.50. Set to 0.25, Set to 20% of the rated power of each heating zone; For heating zones not included in the graded coking risk list For the low-temperature heating zones, compensation power increases are calculated and applied based on the overall heat conservation constraint within the furnace. The conservation constraint is based on the actual power reduction of each risk zone, and its expression is as follows: ; in, Indicates the first Compensation power increase for each low-temperature heating zone; compensation power increase for each low-temperature heating zone. The heat source input power of each zone is allocated according to the proportion of its current input power to the total input power of all low-temperature heating zones, and the adjusted input power of each low-temperature heating zone must not exceed the rated power limit of that zone. When the proportionally allocated compensation power causes the input power of the heat source in a certain low-temperature zone to exceed... At that time, the partition compensation power boost is truncated to The difference between the current heat source input power of the partition and the remaining compensation power generated by the cut-off is proportionally redistributed to the remaining low-temperature heating partitions whose rated power has not yet been saturated. The equation ensures that the total total heat input in the furnace remains unchanged before and after the adjustment.

[0035] It should be noted that the sequence numbers of the above embodiments of the present invention are merely for descriptive purposes and do not represent the superiority or inferiority of the embodiments. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, apparatus, article, or method that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, apparatus, article, or method. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, apparatus, article, or method that includes that element.

[0036] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) as described above, and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0037] The above are merely preferred embodiments of the present invention and do not limit the scope of the patent. Any equivalent structural or procedural transformations made based on the description and drawings of the present invention, or direct or indirect applications in other related technical fields, are similarly included within the scope of patent protection of the present invention.

Claims

1. A resin melting furnace, characterized in that, Including a support (1), a lower end cap (2), a cylinder (3), and an upper end cap (5), characterized in that: The support (1) is fixedly installed with a lower end cap (2). The bottom of the cylinder (3) is sealed to the top of the lower end cap (2). The top of the cylinder (3) is sealed to the bottom of the upper end cap (5) to form a closed melting cavity. The top surface of the upper end cap (5) is provided with flange pipe 2 (9), flange pipe 3 (10), lifting lug (4) and pressure gauge (11). The side of the upper end cap (5) is provided with vent hole 1 (13). The right side wall of the cylinder (3) is provided with nameplate (6) and sight glass (12) from bottom to top. The left side wall of the lower end cap (2) is provided with flange pipe 1 (8). The side wall of the support (1) is fixedly provided with grounding plate (7). The top end face of the upper end cap (5) is reserved with multiple sets of functional interfaces, including discharge port (14), spare port one (15), spare port two (16), safety valve port (17), pressure gauge port (18), vent port (19), sight glass port (20), and temperature gauge connector (21).

2. The resin melting furnace according to claim 1, characterized in that: The lifting lugs (4) are symmetrically welded and fixed to the top two sides of the upper end cap (5), and the base plate (7) is fixedly installed on the side of the support (1).

3. The rosin melting furnace according to claim 1, characterized in that: The sight glass (12) is sealed and assembled on the upper side wall of the cylinder (3), and the sight glass (12) corresponds to the sight glass opening (20) at the top of the upper end cap (5).

4. The resin melting furnace according to claim 1, characterized in that: The pressure gauge (11) is installed inside the pressure gauge port (18) of the upper end cap (5), and the safety valve is installed inside the safety valve port (17).

5. The rosin melting furnace according to claim 1, characterized in that: The temperature sensor is installed inside the temperature gauge connector (21), and the nameplate (6) is fixed to the right side wall of the cylinder (3).

6. A method for preventing resin coking in a resin melting furnace as described in any one of claims 1 to 5, characterized in that, Includes the following steps: S1: Extract the temperature reading of the temperature sensor at the top of the upper head at the current moment, perform solid-liquid interface detection on the furnace image taken by the side wall sight glass of the cylinder, take the solid-liquid phase change temperature of rosin as the physical constraint boundary, take the temperature reading as the global reference anchor point, combine the solid-liquid interface spatial information of the sight glass image to reconstruct the global temperature field, and generate the global high-resolution temperature field of the furnace at the current moment. S2: Will continue The high-resolution temperature field across the entire furnace is arranged chronologically to form a temporal input. Based on the viscosity-temperature characteristics of rosin, the current temperature of each spatial node is converted into a heat transfer modulation weight. This heat transfer modulation weight is then injected into the lateral heat diffusion and propagation process of the viscosity-temperature modulated heat migration prediction network, outputting the future... Predicted sequence of global temperature field within the frame furnace; S3: Extract the temperature peak and continuous temperature rise rate of each spatial node in the predicted temperature field sequence. According to the double-layer coking risk judgment criterion, classify and mark each spatial node as a high-risk coking node or a coking early warning node. Aggregate all marked nodes into coking risk areas according to spatial connectivity. Map the spatial coordinates of each coking risk area to the corresponding heating zone number of the furnace body and output a graded coking risk heating zone list. S4: Implement differentiated heat source input power regulation for each heating zone based on the coking risk level. For low-temperature heating zones not included in the coking risk list, proportionally compensate and increase the heat source input power to meet the overall heat conservation constraint in the furnace and achieve uniform heating and anti-coking temperature control in the furnace.

7. The temperature control method for preventing resin scorching according to claim 6, characterized in that, Step S1 includes: Extract the temperature sensor installed inside the temperature gauge connector on the top of the upper end cap at the current moment. The temperature reading is recorded as the global reference temperature. Simultaneously, an image acquisition device installed at the sight glass on the side wall of the cylinder acquires images at any time. The captured images inside the furnace are input into a phase boundary recognition network composed of a convolutional feature extraction layer and a boundary contour regression layer, and the solid-liquid interface contour is output. Based on the pre-defined projection mapping relationship between the sight glass image coordinate system and the furnace three-dimensional space coordinate system, the solid-liquid interface contour is transformed from the image coordinate system to the furnace three-dimensional space coordinate system, thereby determining the set of spatial nodes covered by the visible area of ​​the sight glass in the furnace three-dimensional space. ;right Within each spatial node, nodes within the area enclosed by the solid-liquid interface contour are labeled as resin solid phase nodes, and nodes outside the contour are labeled as resin liquid phase nodes; for nodes not belonging to... The furnace interior space nodes, based on the global reference temperature Solid-liquid phase transition temperature of rosin Phase determination is performed based on the magnitude relationship: when Below When all non-visible spatial nodes are labeled as resin solid phase nodes; when Not less than At that time, all non-visible spatial nodes were labeled as resin liquid phase nodes; all The set of all solid-phase nodes in a given spatial node is denoted as the resin solid-phase region set. Let the set of all liquid phase nodes be denoted as the set of rosin liquid phase regions. ; Based on the solid-liquid phase transition temperature of rosin As a physical constraint boundary, for the entire furnace area Phase constraints are imposed on the reconstruction temperature of each spatial node: solid-phase region set The upper limit of the reconstruction temperature of each spatial node does not exceed Liquid phase region collection The lower bound of the reconstruction temperature of each spatial node is not lower than Under the condition that the phase state constraint is satisfied, the global reference temperature is... With the entire furnace area The phase state label vectors of each spatial node are jointly input into the phase state-guided denser network, which consists of three layers of spatial convolution and bilinear interpolation upsampling stacked alternately, extending the temperature readings and phase state spatial distribution information to the entire furnace domain. Each spatial node; the phase constraint is directly applied to the output of each layer of the phase-guided denser network by a truncation operation, so that the reconstruction temperature of each node satisfies: ; in, Indicates time No. The reconstruction temperature of each spatial node. , This indicates that the phase-guided denser network is for the first... Each spatial node at time The output is the unconstrained temperature prediction. This indicates the solid-liquid phase transition temperature of rosin. Take the smaller value for the two parameters. Take the larger value of the two parameters; Reconstruction temperature of each spatial node Overall composition moment High-resolution temperature field throughout the furnace .

8. The temperature control method for preventing resin scorching according to claim 7, characterized in that, Step S2 includes: Time Previous consecutive Full-range high-resolution temperature field within the frame furnace Arranged in chronological order, they form a temporal input tensor, where To preset the number of historical frames, the temporal input tensor is input into the viscosity-temperature modulation thermal migration prediction network. The network is composed of alternating stacks of spatial convolution modules that perform local receptive field convolution operations along the spatial dimension and temporal recursive modules that perform gated state updates along the temporal dimension. Based on the viscosity-temperature characteristic that the viscosity of rosin decreases monotonically with increasing local temperature, each spatial node... At any moment Reconstruction temperature Convert to heat transfer modulation weight The heat transfer modulation weight quantifies the modulation effect of the local flowability of rosin on the intensity of lateral heat diffusion, and the calculation formula is as follows: ; in, Indicates the first Each spatial node at time Heat transfer modulation weight, Indicates the first Each spatial node at time The reconstruction temperature This indicates the solid-liquid phase transition temperature of rosin. Indicates the upper limit of the furnace temperature. This is an indicator function that takes the value 1 when the condition inside the parentheses is true and 0 when it is false. The range of values ​​is When the node is in the solid phase region Lateral heat diffusion is suppressed; when the node temperature is higher than hour, It increases with increasing temperature, and its ability to diffuse heat laterally is enhanced. The heat transfer modulation weight As the channel modulation coefficients of each convolution kernel in the spatial convolution module of the viscosity-temperature modulated heat migration prediction network, they are multiplied element-wise with the current hidden layer feature vector of each spatial node and then participate in the aggregation of convolutional information between adjacent nodes, enabling the network to accurately predict lateral heat migration. Apply strong diffusion weights to value nodes and apply them to low-value nodes. A weak diffusion weight is applied to the value nodes to automatically distinguish between heat accumulation trend nodes and heat diffusion trend nodes, and the future value is output. Prediction sequence of global temperature field in frame furnace ,in This is the preset number of prediction frames.

9. The temperature control method for preventing resin scorching according to claim 8, characterized in that, Step S3 includes: For the future Prediction sequence of global temperature field in frame furnace Each spatial node Extracting peak temperature With sustained rate of temperature rise Two characteristics: the peak temperature For the first Each spatial node The maximum temperature value within the frame prediction sequence; the sustained rate of temperature rise. For the first Each spatial node The slope of the linear fit of temperature change over time within the frame prediction sequence. A positive slope indicates a continuous temperature rise trend, while a zero or negative slope indicates a stable or falling temperature trend. Based on the two-layer coking risk assessment criteria, each spatial node is classified and labeled: the first layer is determined by the critical temperature of resin coking. The threshold for determining the peak temperature. Spatial nodes are directly marked as high-risk nodes for coking; the second layer of judgment is based on the warning temperature threshold. The threshold for determining the peak temperature and the threshold for the early warning temperature rise rate. The threshold for rate determination is denoted as follows: , and The spatial nodes are marked as coking warning nodes; the remaining spatial nodes are marked as normal nodes. All high-risk coking nodes and coking early warning nodes are aggregated into a set of high-risk coking areas based on three-dimensional spatial connectivity. and coking warning area set ;Will The three-dimensional spatial coordinates of nodes within each region are matched with a pre-established heating zone spatial index table. The covered heating zone number is written into the high-risk heating zone list. ;Will The covered heating zone number is written into the warning heating zone list. Output a list of heating zones with graded coking risk. .

10. The temperature control method for preventing resin scorching according to claim 9, characterized in that, Step S4 includes: List of high-risk heating zones In each heating zone, power is reduced according to the proportion of high-risk zones. The input power of the partition heat source is changed by Reduce to ; List of heating zones for early warning Each heating zone in the middle will reduce power according to the warning ratio. The input power of the partition heat source is changed by Reduce to ,in Indicates the first Current heat source input power for each partition After implementing the power reduction regulation, verify whether the heat source input power after regulation in each zone is lower than the minimum power limit required to maintain the continuous dissolution of pine resin. For power levels lower than 1 after regulation The partition limits the heat source input power of the partition to The actual power reduction amount in each risk zone Defined as the current heat source input power of the partition. The difference between the final heat source input power after clamping and the actual power input after clamping, i.e. ,in The power reduction ratio corresponding to the partition is taken as follows: high-risk heating partition is... Early warning heating zone ; For heating zones not included in the graded coking risk list For the low-temperature heating zones, compensation power increases are calculated and applied based on the overall heat conservation constraint within the furnace. The conservation constraint is based on the actual power reduction of each risk zone, and its expression is as follows: ; in, Indicates the first Compensation power increase for each low-temperature heating zone; compensation power increase for each low-temperature heating zone. The heat source input power of each zone is allocated according to the proportion of its current input power to the total input power of all low-temperature heating zones, and the adjusted input power of each low-temperature heating zone must not exceed the rated power limit of that zone. When the proportionally allocated compensation power causes the input power of the heat source in a certain low-temperature zone to exceed... At that time, the partition compensation power boost is truncated to The difference between the current heat source input power of the partition and the remaining compensation power generated by the cut-off is proportionally redistributed to the remaining low-temperature heating partitions whose rated power has not yet been saturated. The equation ensures that the total total heat input in the furnace remains unchanged before and after the adjustment.