Carbide slag drying temperature control method and system based on multi-agent cooperation

Abstract

The invention discloses a carbide slag drying temperature control method and system based on multi-agent collaboration, and relates to the technical field of industrial heat treatment.The method comprises the steps that S1, real-time pressure data and temperature data are collected through a sensor network deployed in a rock slag accumulation area, and a current bearing distribution diagram and a current temperature distribution diagram of rock slag accumulation are obtained; s2, according to the bearing distribution diagram and the temperature distribution diagram, a convolutional neural network model is adopted to process diagram data, and the structural instability risk level in rock ballast accumulation is determined; s3, if the structural instability risk level is higher than a preset threshold value, reducing temperature input of a corresponding area by adjusting power parameters of a heating mode to obtain an optimized heating distribution scheme; according to the carbide slag drying temperature control method and system based on multi-agent collaboration, through multi-model collaboration and dynamic feedback, the stacking safety and drying efficiency of the carbide slag are remarkably improved, and the instability and overheating risks are reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of industrial heat treatment, and in particular to a carbide slag drying temperature control method and system based on multi-agent collaboration. Background Art

[0002] In industrial production and resource processing, carbide slag drying is a key technology. Slag drying not only impacts subsequent resource utilization but also directly impacts production safety and equipment lifespan. An efficient drying system must meet temperature requirements while ensuring the stability of the carbide slag stack structure to prevent collapse or equipment damage caused by improper operation. Technological advancements in this area are crucial for improving production efficiency and reducing safety risks. However, current drying technologies still face numerous challenges in balancing temperature control and structural stability, and innovative breakthroughs are urgently needed.

[0003] Existing drying systems often rely on a single heating method or static structural monitoring, making them ill-suited to the complex characteristics of carbide slag piles. The shape and load-bearing distribution of carbide slag piles dynamically change depending on the material type, particle size, and stacking method. Traditional methods often overlook this dynamic nature, resulting in an inability to accurately respond to real-time changes in the pile structure during the heating process. Furthermore, existing systems often lack comprehensive consideration of structural stability in temperature control, and lack effective coordination between heating method adjustments and structural status monitoring. This makes the drying process susceptible to localized overheating or structural instability, impacting production safety and efficiency. The core technical challenge in carbide slag drying lies in achieving a dynamic balance between temperature control and structural stability. The primary technical factor is the highly heterogeneous load-bearing distribution of carbide slag piles. Differences in particle size, moisture content, or bulk density result in uneven load-bearing capacity in different areas of carbide slag, potentially leading to the risk of slippage or collapse due to concentrated forces in certain areas. This heterogeneous load-bearing distribution further introduces a second technical factor: the direct impact of temperature changes on structural stability. Rapid or uneven heating can cause rapid evaporation of water within carbide slag, leading to volume expansion or contraction, which in turn changes the load distribution and increases the risk of structural instability. These two intertwined factors require the drying system to dynamically monitor the load distribution while precisely controlling the heating method to prevent temperature fluctuations from damaging the structure.

[0004] Specifically, in actual production, the drying process of carbide slag piles often faces the following problem: when the heating system increases the temperature to accelerate the evaporation of water, cracks may occur inside the carbide slag pile due to local overheating, resulting in a decrease in bearing capacity and even causing the pile to slip. Summary of the Invention

[0005] The purpose of the present invention is to provide a temperature control method and system for carbide slag drying based on multi-agent collaboration, which can accurately control the heating mode under dynamically changing load distribution to avoid damage to the carbide slag stacking structure caused by temperature changes.

[0006] To achieve the above-mentioned purpose, the present invention provides the following technical solutions: a temperature control method for carbide slag drying based on multi-agent collaboration, the method comprising: S1, collecting real-time pressure data and temperature data through a sensor network deployed in the carbide slag accumulation area, and obtaining the current load-bearing distribution map and temperature distribution map of the carbide slag accumulation; S2, using a convolutional neural network model to process the graph data according to the load-bearing distribution map and the temperature distribution map, and determining the structural instability risk level in the carbide slag accumulation; S3, if the structural instability risk level is higher than a preset threshold, reducing the temperature input of the corresponding area by adjusting the power parameters of the heating method, and obtaining an optimized heating distribution scheme; S4, obtaining the optimized After determining the heating distribution scheme, the support vector machine model is used to analyze the impact of the scheme on the overall drying efficiency to determine whether it meets the efficiency requirements under the production safety standards; S5. If the efficiency requirements under the production safety standards are met, the heating method execution scheme is adjusted in real time to obtain the updated carbide slag accumulation temperature field and load-bearing distribution field; S6. Based on the updated carbide slag accumulation temperature field and load-bearing distribution field, the recurrent neural network model is used to predict the structural stability change trend in the future time and determine the potential local thermal evolution path; S7. Through the potential local thermal evolution path, a targeted heating control instruction sequence is generated to obtain a dynamic monitoring cycle mechanism to maintain structural stability.

[0007] Preferably, S1 includes grid deployment in the carbide slag accumulation area through a sensor network to obtain pressure data and temperature data and generate an initial data set; based on the initial data set, a data cleaning method is used to remove noise data to obtain cleaned pressure data and temperature data; based on the cleaned pressure data, an interpolation algorithm is used to calculate the load-bearing distribution to generate a load-bearing distribution map of the carbide slag accumulation; based on the cleaned temperature data, an interpolation algorithm is used to calculate the temperature distribution to generate a temperature distribution map of the carbide slag accumulation; if there is an area in the load-bearing distribution map where the pressure anomaly is higher than a preset threshold, the regional data is re-collected through the sensor network to obtain updated pressure data; if there is an area in the temperature distribution map where the temperature anomaly is higher than the preset threshold, the regional data is re-collected through the sensor network to obtain updated temperature data; based on the updated pressure data and temperature data, a weighted average algorithm is used to fuse multiple sets of data to generate an optimized load-bearing distribution map and temperature distribution map.

[0008] Preferably, S2 includes obtaining the load-bearing distribution map and temperature distribution map of the carbide slag accumulation area through a sensor network to generate initial image data; using a data preprocessing method to standardize the initial image data to obtain standardized image data; using a convolutional neural network to extract features of the standardized image data to obtain a feature vector; based on the feature vector, using a convolutional neural network to calculate the probability of occurrence of a local overheating area to obtain a probability distribution; if there is an area in the probability distribution that is higher than a preset threshold, the load-bearing and temperature data of the area are re-collected through the sensor network to obtain updated image data; using a convolutional neural network to analyze the updated image data, calculate the structural instability risk level, and obtain a risk assessment result; based on the risk assessment result, using a regional division method to grade and label the carbide slag accumulation area to obtain a graded distribution map.

[0009] Preferably, S3 includes obtaining temperature data and load-bearing data of the calcium carbide slag accumulation area through a sensor network to generate an initial distribution data set; processing the initial distribution data set using a data standardization method to obtain a standardized data set; extracting features from the standardized data set through a convolutional neural network to obtain a feature vector; if there is a risk indicator higher than a preset threshold in the feature vector, re-collecting the temperature data and load-bearing data of the corresponding area through the sensor network to generate an updated distribution data set; analyzing the updated distribution data set using a logistic regression model to determine the risk level of structural instability; adjusting the heating power parameters according to the risk level of structural instability to generate an optimized heating distribution scheme; verifying the optimized heating distribution scheme through the sensor network, obtaining a verification data set, and determining the stability of the temperature and load-bearing data.

[0010] Preferably, the S4 includes obtaining the temperature control parameters and load-bearing distribution data corresponding to the optimized heating distribution scheme through the sensor network to obtain an initial efficiency data set; processing the initial efficiency data set using a data standardization method to obtain a standardized efficiency data set; classifying and analyzing the standardized efficiency data set through a support vector machine model to obtain drying efficiency data; if the drying efficiency data is lower than the efficiency requirements preset by the production safety standards, re-collecting the temperature control parameters and load-bearing distribution data through the sensor network to obtain an updated efficiency data set; analyzing the updated efficiency data set using a logistic regression model to determine risk assessment indicators; adjusting the temperature control parameters according to the risk assessment indicators to obtain an optimized adjustment strategy; verifying the optimized adjustment strategy through the sensor network, obtaining a verification data set, and determining whether the drying efficiency data meets the production safety standards.

[0011] Preferably, the S5 includes obtaining updated carbide slag accumulation temperature field data and load-bearing distribution field data through a sensor network to obtain an initial field distribution data set; processing the initial field distribution data set using a data standardization method to obtain a standardized field distribution data set; classifying and analyzing the standardized field distribution data set through a random forest model to obtain feature classification results of the temperature field and load-bearing distribution; if the feature classification results show that the temperature field or load-bearing distribution deviates from a preset stability threshold, re-collecting the temperature field data and load-bearing distribution field data through the sensor network to obtain an updated field distribution data set; using the principal component analysis method to perform dimensionality reduction processing on the updated field distribution data set to obtain a reduced dimensionality feature data set; based on the reduced dimensionality feature data set, analyzing the stability of the temperature field and load-bearing distribution through a logistic regression model to obtain a stability evaluation result; if the stability evaluation result meets the production safety standard, adjusting the heating distribution parameters in real time through the sensor network to obtain optimized temperature field data and load-bearing distribution field data.

[0012] Preferably, S6 includes obtaining temperature field data and load-bearing distribution field data of carbide slag accumulation through a sensor network to obtain a real-time distribution data set; processing the real-time distribution data set using a data standardization method to obtain a standardized distribution data set; analyzing the standardized distribution data set through a recurrent neural network model to predict the structural stability change trend and obtain a stability trend prediction result; if the stability trend prediction result shows that the local thermal risk is higher than the preset threshold, re-collecting the temperature field data through the sensor network to obtain an updated temperature data set; using a clustering analysis method to group the updated temperature data set to determine the regional distribution where the local overheating risk is higher than the preset threshold to obtain a hot area data set; based on the hot area data set, judging the local overheating path through a preset threshold to obtain a thermal evolution path distribution; optimizing the thermal evolution path distribution by adjusting the heating distribution parameters to obtain optimized temperature field data and load-bearing distribution field data.

[0013] Preferably, S7 includes acquiring temperature distribution data of local overheating areas through a sensor network to obtain a real-time temperature data set; grouping the temperature distribution according to the real-time temperature data set using a K-means clustering analysis method to determine the regional distribution where the local temperature is higher than a preset threshold value to obtain an overheating area data set; judging the temperature anomaly points in the overheating area data set through a preset threshold value to generate an initial heating control instruction sequence to obtain a control instruction data set; analyzing the control instruction data set using a long short-term memory network model to predict the structural stability change trend to obtain a stability trend prediction result.

[0014] Preferably, S7 also includes if the stability trend prediction result shows that the local thermal risk is higher than a preset threshold, re-collecting the temperature distribution data and the load-bearing distribution data through the sensor network to obtain an updated distribution data set; based on the updated distribution data set, using the support vector machine model to classify the evolution path of the overheating area to obtain an optimized overheating evolution path distribution; through the optimized overheating evolution path distribution, generating a dynamically adjusted heating control instruction sequence to determine the real-time monitoring cycle mechanism.

[0015] A carbide slag drying temperature control system based on multi-agent collaboration is used to implement the steps of the carbide slag drying temperature control method based on multi-agent collaboration. The system includes a sensor network module, which collects real-time pressure data and temperature data through a sensor network deployed in the carbide slag accumulation area to obtain the current load-bearing distribution map and temperature distribution map of the carbide slag accumulation; an image processing module, which uses a convolutional neural network model to process the image data according to the load-bearing distribution map and the temperature distribution map to determine the structural instability risk level in the carbide slag accumulation; a heating control module, which reduces the temperature input of the corresponding area by adjusting the power parameters of the heating method if the structural instability risk level is higher than a preset threshold, and obtains an optimized heating distribution scheme; an efficiency evaluation module Block, after obtaining the optimized heating distribution plan, uses the support vector machine model to analyze the impact of the plan on the overall drying efficiency to determine whether it meets the efficiency requirements under the production safety standards; the execution control module, if the efficiency requirements under the production safety standards are met, adjusts the heating method execution plan in real time to obtain the updated calcium carbide slag accumulation temperature field and load-bearing distribution field; the stability prediction module, based on the updated calcium carbide slag accumulation temperature field and load-bearing distribution field, uses the recurrent neural network model to predict the structural stability change trend in the future time and determine the potential local thermal evolution path; the instruction generation module, through the potential local thermal evolution path, generates a targeted heating control instruction sequence to obtain a dynamic monitoring cycle mechanism to maintain structural stability.

[0016] It can be seen from the above technical solution that the present invention has the following beneficial effects: The carbide slag drying temperature control method and system based on multi-agent collaboration targets the business scenario problems of local overheating and structural instability in carbide slag accumulation. It collects pressure and temperature data in real time through a sensor network, generates a load-bearing distribution map and a temperature distribution map, and uses a convolutional neural network to analyze the probability of local overheating and the risk of structural instability. If the risk exceeds the threshold, the present invention optimizes the temperature input by adjusting the heating power, generates a heating distribution plan, and uses a support vector machine to evaluate its impact on the drying efficiency to ensure that production safety standards are met. Subsequently, the present invention executes the optimization plan in real time, updates the temperature field and the load-bearing distribution field, combines the recurrent neural network to predict the structural stability trend and the overheating evolution path, generates a dynamic control instruction sequence, and realizes continuous monitoring and control. The present invention significantly improves the safety and drying efficiency of carbide slag accumulation and reduces the risk of instability and overheating through multi-model collaboration and dynamic feedback. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 This is a flow chart of the carbide slag drying temperature control method based on multi-agent collaboration of the present invention; Figure 2 This is a module connection diagram of the carbide slag drying temperature control system based on multi-agent collaboration of the present invention. DETAILED DESCRIPTION

[0018] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0019] like Figure 1As shown, the present invention provides a technical solution: a temperature control method for carbide slag drying based on multi-agent collaboration, the method comprising S1, collecting real-time pressure data and temperature data through a sensor network deployed in the carbide slag accumulation area, and obtaining the current load-bearing distribution map and temperature distribution map of the carbide slag accumulation; S2, using a convolutional neural network model to process the graph data according to the load-bearing distribution map and the temperature distribution map, and determining the structural instability risk level in the carbide slag accumulation; S3, if the structural instability risk level is higher than a preset threshold, reducing the temperature input of the corresponding area by adjusting the power parameters of the heating method, and obtaining an optimized heating distribution scheme; S4, obtaining the optimized heating After the distribution plan is formulated, the support vector machine model is used to analyze the impact of the plan on the overall drying efficiency to determine whether it meets the efficiency requirements under the production safety standards; S5. If the efficiency requirements under the production safety standards are met, the heating method is adjusted in real time to execute the plan to obtain the updated carbide slag accumulation temperature field and load-bearing distribution field; S6. Based on the updated carbide slag accumulation temperature field and load-bearing distribution field, the recurrent neural network model is used to predict the structural stability change trend in the future time and determine the potential local thermal evolution path; S7. Through the potential local thermal evolution path, a targeted heating control instruction sequence is generated to obtain a dynamic monitoring cycle mechanism to maintain structural stability.

[0020] This implementation targets structural stability during the drying process of carbide slag piles. Using a sensor network, pressure and temperature sensors are deployed throughout the carbide slag accumulation area to achieve real-time sensing of the accumulation state. The collected temperature and load-bearing information is then used to construct a two-dimensional or three-dimensional distribution map, which is then fed into a trained convolutional neural network model to identify high-risk areas and the overall structural stability level. When the system determines that a potential instability risk exists and exceeds a safety threshold, it initiates a local temperature control optimization program, dynamically adjusting the power output of the heating unit to reduce heat input to the corresponding high-risk area. The system then uses a support vector machine model to evaluate the impact of this heating adjustment strategy on overall drying efficiency, ensuring a balance between drying rate and safety. If the strategy is determined to meet safety standards, the adjusted temperature control strategy is applied in real time, updating the thermodynamic state of the accumulation area. Based on this, a recurrent neural network predicts future structural changes based on the updated temperature and load-bearing data, identifying potential local thermal evolution paths and providing decision-making support for subsequent heating control, forming a complete closed-loop control chain. Finally, based on the prediction results, the system generates an instruction sequence for fine-grained control of the heating units, ensuring structural stability and dynamic optimization of drying results during the drying process.

[0021] Compared to traditional single-point heating or fixed-strategy heating systems, this implementation significantly improves the structural safety and drying efficiency of the carbide slag drying process. By introducing multi-agent model collaborative control and machine learning algorithms, adaptive temperature control and risk perception are achieved during the drying process, effectively avoiding structural collapse due to local overheating or decreased drying efficiency due to uneven heat distribution. Furthermore, by establishing a dynamic monitoring cycle mechanism, the system can adjust its control strategy in real time to respond to on-site changes, providing greater robustness and environmental adaptability.

[0022] S1 includes grid deployment in the carbide slag accumulation area through a sensor network to obtain pressure data and temperature data and generate an initial data set; based on the initial data set, a data cleaning method is used to remove noise data to obtain cleaned pressure data and temperature data; based on the cleaned pressure data, an interpolation algorithm is used to calculate the load-bearing distribution to generate a load-bearing distribution map of the carbide slag accumulation; based on the cleaned temperature data, an interpolation algorithm is used to calculate the temperature distribution to generate a temperature distribution map of the carbide slag accumulation; if there is an area in the load-bearing distribution map where the pressure anomaly is higher than a preset threshold, the regional data is re-collected through the sensor network to obtain updated pressure data; if there is an area in the temperature distribution map where the temperature anomaly is higher than the preset threshold, the regional data is re-collected through the sensor network to obtain updated temperature data; based on the updated pressure data and temperature data, a weighted average algorithm is used to fuse multiple sets of data to generate an optimized load-bearing distribution map and temperature distribution map.

[0023] During implementation, the carbide slag accumulation area is first gridded using a sensor network. Specifically, the accumulation area is divided into several regular cell areas, with a set of pressure and temperature sensors placed at the center of each cell. The grid density is determined based on the overall area and structural complexity of the accumulation area. For example, a 1,000-square-meter accumulation area is typically divided into 100 cells, each measuring 10 square meters, ensuring comprehensive data coverage and high spatial resolution. Each sensor collects data every five minutes, forming an initial pressure and temperature dataset containing the current pressure and temperature values ​​at each grid location.

[0024] The collected initial data set then enters the data cleaning step. The specific method for data cleaning is to first set the normal variation range of each type of sensor data. For example, the normal range of pressure values ​​is 0 to 500 kilopascals, and the normal range of temperature values ​​is 0 to 300 degrees Celsius. If a data point falls outside this range, it is marked as abnormal data and directly eliminated. The remaining data is then median smoothed. This means that each data point is sorted with the values ​​of the eight surrounding grid points, and the median value is taken as the cleaned value of the point. This eliminates occasional spike interference and obtains the cleaned pressure and temperature data.

[0025] After obtaining the post-cleaning pressure data, an interpolation algorithm is used to calculate the load-bearing distribution map. The interpolation process involves selecting the four closest sensor data points around any point without a sensor, with that point as the center. The pressure values ​​are then weighted based on the distance between each point and the target point. Weights are assigned based on the principle that closer distances give higher weights, while farther distances give lower weights. The specific weights are determined inversely, proportional to the inverse of the distance. The value of each calculated point is the estimated pressure value at that location, and the calculated values ​​of all points constitute the load-bearing distribution map of the carbide slag accumulation.

[0026] The generation process of temperature distribution map and load distribution Figure 1 The same four-point weighted interpolation method is used, with the data source being the cleaned temperature data. For any temperature value at a non-sensor location, a weight is calculated based on its distance from the four surrounding sensors. The weighted average is then used to calculate the temperature value at the target location, thus forming a complete temperature distribution map.

[0027] Subsequently, the load-bearing distribution map and the temperature distribution map were separately checked for anomalies. The degree of anomaly was determined based on the following criteria: if the pressure value in a certain area of ​​the load-bearing distribution map was more than 50 percent higher than the surrounding average, the area was considered to have a pressure anomaly; in the temperature distribution map, if the temperature in a certain area was more than 30 degrees Celsius higher than the surrounding average, the area was considered to have a temperature anomaly. These percentages and temperature differences represent the preset anomaly thresholds, which are statistical averages derived from historical accumulation collapse case studies. This ensures that potential anomalies can be detected while avoiding overreaction.

[0028] Once an abnormal area is identified, the sensor network is immediately instructed to resample the area at a higher frequency. This resampling can be done by deploying a higher density of sensors within the marked abnormal area or by increasing the sampling frequency of existing sensors to once per minute for a period of ten minutes, collecting new pressure and temperature data.

[0029] After obtaining the updated regional data, a weighted average algorithm is used to fuse the original cleaned data with the updated data. Specifically, the original and updated values ​​at the same location are weighted differently, with the updated data weighted at 70% and the original value at 30%. The final optimized values ​​are calculated based on the weighted ratios. This method is used to process the data from all updated areas, resulting in the final optimized load-bearing and temperature distribution maps.

[0030] S2 includes obtaining the load-bearing distribution map and temperature distribution map of the carbide slag accumulation area through the sensor network to generate initial image data; using the data preprocessing method to standardize the initial image data to obtain standardized image data; using the convolutional neural network to extract features of the standardized image data to obtain a feature vector; based on the feature vector, using the convolutional neural network to calculate the occurrence probability of the local overheating area to obtain a probability distribution; if there is an area above the preset threshold in the probability distribution, the load-bearing and temperature data of the area are re-collected through the sensor network to obtain updated image data; using the convolutional neural network to analyze the updated image data, calculate the structural instability risk level, and obtain a risk assessment result; based on the risk assessment result, the area division method is used to grade and label the carbide slag accumulation area to obtain a graded distribution map.

[0031] In this embodiment, pressure and temperature data are first collected through a sensor network deployed in the carbide slag accumulation area, and a two-dimensional grid structure is constructed based on the spatial positions corresponding to the sensors. Each grid unit represents a sensor measurement point at a spatial position. The load-bearing distribution map and temperature distribution map obtained above are used to encode them in the form of an image into a two-dimensional matrix, where the load-bearing value and the temperature value are respectively used as two channel information of the image to form the initial image data. The resolution of the image data is consistent with the grid division, usually set to one pixel per meter, and the specific image size depends on the actual area of ​​the accumulation area. For example, when the accumulation area is a rectangular area with a length of 20 meters and a width of 10 meters, the image data size is 20 times 10 pixels.

[0032] The initial image data is first standardized before being input into the neural network. The standardization method is linear normalization, that is, the load-bearing value and temperature value of each pixel are linearly compressed according to the preset maximum and minimum values, so that all pixel values ​​are mapped to floating-point numbers between 0 and 1. Specifically, the maximum value of the load-bearing value is set to 500 kPa, and the minimum value is 0 kPa; the maximum value of the temperature value is set to 300 degrees Celsius, and the minimum value is 0 degrees Celsius. The standardized image data is obtained by subtracting the minimum value from each pixel value and dividing it by the difference between the maximum and minimum values. This standardization step is used to eliminate the difference in data magnitude between different regions and ensure that the neural network has good numerical stability during the processing process.

[0033] Normalized image data is input into a convolutional neural network for feature extraction. The network structure consists of three convolutional layers and two pooling layers. The convolution kernel size is fixed at 3x3 with a stride of 1. Pooling uses a 2x2 max pooling method, and the activation function uniformly uses the rectified linear unit function. This network structure extracts local features from the input image, outputting a feature vector of length 256, representing the overall characteristics of the temperature distribution and load-bearing structure in the image. This feature vector is input into the subsequent classification structure, a computational module consisting of a fully connected layer and a softmax layer, which calculates the probability of each grid point being a local overheating area. The network's final output is a two-dimensional probability map that matches the input image size, indicating the probability of each location being a local overheating area.

[0034] The threshold for determining local overheating is set at 0.8. This means that when the probability of overheating at a grid point exceeds 0.8, that point is considered to be at serious overheating risk. This threshold is determined based on historical accumulated accident data and field experience, and has high sensitivity and a low false positive rate. Once an area with a value above the threshold is identified, the sensors within the corresponding grid area are immediately instructed to enter high-frequency sampling mode, sampling once per minute and continuously collecting the latest pressure and temperature data for 5 minutes.

[0035] The collected new data is re-normalized in the same way, converted into updated image data, and re-entered into the original convolutional neural network for risk analysis. After the network performs the aforementioned feature extraction and risk assessment process on the updated image data, it outputs the structural instability risk level. The risk level is defined as a continuous value between 0 and 1, and is divided into 5 risk level intervals based on this value, among which 0 to 0.2 is low risk, 0.2 to 0.4 is medium-low risk, 0.4 to 0.6 is medium risk, 0.6 to 0.8 is medium-high risk, and 0.8 to 1 is high risk. The risk level classification standard is based on the carbide slag accumulation stability assessment model, and is comprehensively evaluated based on multiple characteristics such as regional thermal intensity gradient and local structural force concentration to ensure that the grade results are accurate and have guiding value.

[0036] Based on these risk level results, a regional division method was implemented. This method marked the corresponding grid points in the original image with different colors according to their risk level. The colors were set as blue for low risk, green for medium-low risk, yellow for medium risk, orange for medium-high risk, and red for high risk, forming a final graded distribution map. This image served as the basic input for subsequent temperature control adjustments. Based on the level of each risk zone, a corresponding heating power limitation strategy was implemented to ensure the overall safety and stability of the stacking structure.

[0037] S3 includes obtaining temperature data and load-bearing data of the carbide slag accumulation area through a sensor network to generate an initial distribution data set; using a data standardization method to process the initial distribution data set to obtain a standardized data set; extracting features from the standardized data set through a convolutional neural network to obtain a feature vector; if there is a risk indicator higher than a preset threshold in the feature vector, re-collecting the temperature data and load-bearing data of the corresponding area through the sensor network to generate an updated distribution data set; using a logistic regression model to analyze the updated distribution data set to determine the risk level of structural instability; adjusting the heating power parameters according to the risk level of structural instability to generate an optimized heating distribution plan; verifying the optimized heating distribution plan through the sensor network, obtaining a verification data set, and determining the stability of the temperature and load-bearing data.

[0038] In this embodiment, a sensor network deployed in the carbide slag accumulation area first collects temperature and load-bearing data for each grid unit. The temperature sensor is installed at a density of one per square meter, and the pressure sensor is installed at a density of one per two square meters, ensuring comprehensive and dense data coverage. All sensors sample once every five minutes. After a complete sampling cycle, the data collected by all sensors are organized uniformly by grid position to construct an initial distribution data set, where each piece of data corresponds to a grid position and includes two attributes: temperature value and load-bearing value. This initial distribution data set is then standardized using linear normalization. Specifically, the temperature value is first set to a standardized range of 0 to 300 degrees Celsius. Each temperature value is subtracted from 0 degrees Celsius and then divided by 300 to obtain a floating-point number between 0 and 1 as the standardized result. The load-bearing value is standardized to a range of 0 to 500 kPa. The processing method is to subtract 0 from each load-bearing value and then divide it by 500, also obtaining a standardized result between 0 and 1. This process ensures that all data maintain a unified numerical scale and eliminates interference caused by different physical units.

[0039] Next, the standardized dataset was input into a trained convolutional neural network model. This neural network model consists of three convolutional layers, each of which uses a 3x3 convolution kernel for sliding calculations to extract spatial features. Each convolution kernel performs a weighted accumulation operation on the input data within its range with a fixed step size. The calculated local features are then processed by an activation function and output to the next layer. The convolutional layers are followed by two maximum pooling layers, each of which uses a 2x2 sliding window to retain the maximum value within the region, thereby reducing the data dimension while retaining key features. The neural network ultimately outputs a feature vector of length 256, which represents the coupled characteristics of the structural stress and thermal distribution of the entire stacking area at the current moment.

[0040] This feature vector is analyzed item by item, containing several risk indicators representing the risk of structural anomaly. If the value of any indicator exceeds the set risk threshold of 0.7, the risk area re-collection mechanism is triggered. This threshold of 0.7 is derived from statistical analysis of the neural network output results of hundreds of accumulation anomaly cases and represents the critical probability of structural instability. At this time, the sensor network is instructed to re-sample the grid area corresponding to the risk indicator at a high frequency, adjusting the sampling frequency to once every minute and the sampling duration to 5 minutes to ensure the latest and most intensive data updates. After the collection is completed, an updated distributed dataset is formed, whose structure is consistent with the initial distributed dataset, but with a higher sampling frequency.

[0041] The updated distribution dataset, after undergoing the same normalization process, was then input into an established logistic regression model. The model used the updated standardized temperature and load values ​​as independent variables to calculate the probability of structural instability. The model internally performs an inner product operation on the input vector and the model parameters based on the trained weights. The output is then mapped to a continuous value between 0 and 1 using a sigmoid function. This value represents the current risk level of structural instability. First, upon identifying a risk indicator above a preset threshold in the feature vector, the sensor network is instructed to re-collect temperature and load data for the corresponding area and its adjacent grids once per minute for five minutes, generating the updated distribution dataset. Each data entry contains two fields: the temperature value (in degrees Celsius) and the load value (in kilopascals) at that grid location. After acquisition, the dataset is normalized by dividing the temperature value by its maximum set value of 300 degrees Celsius and the load value by its maximum set value of 500 kilopascals. These values, respectively, are normalized to obtain normalized temperature and load values ​​between 0 and 1, forming the updated distribution dataset. This normalized dataset is then fed into the logistic regression model one entry at a time. The logistic regression model is a binomial probability model consisting of two parts: a linear combiner and a probability mapping function. For each data point, the model first multiplies the standardized temperature and load-bearing values ​​by the corresponding weight parameters in the logistic regression model. The two products are then summed and a constant bias term is added to form a linear combination value. This value is input as an independent variable into the Sigmoid function, which converts it into a continuous value between 0 and 1. This value is the structural instability risk level corresponding to that data point. The Sigmoid function is a probability mapping function commonly used in logistic regression models. Its function is to compress any real-number input into a numerical range between 0 and 1, thereby converting the model output into a probabilistic result. In specific applications, the input standardized temperature and load-bearing values ​​are first weighted and summed with the weight coefficients obtained from model training. The bias term is then added to form a linear combination value. This linear combination value serves as the input to the Sigmoid function, which, after nonlinear mapping, outputs a floating-point number between 0 and 1. This value represents the probability assessment of whether the current input data falls into the risk category of "structural instability" or "insufficient drying efficiency." The Sigmoid function exhibits distinct monotonicity and an S-shaped curve. When the input value is large, its output approaches 1, indicating that the risk is highly likely to occur. When the input value is small, the output approaches 0, indicating that the risk is extremely low. When the input value is close to 0, the output is 0.5, indicating no significant bias in predicting whether the risk will occur. This risk level represents the probability of structural instability in the grid area under the current temperature and load-bearing conditions.The risk level of each point is divided into risk intervals: when the risk value is less than or equal to 0.2, it is judged as low risk; greater than 0.2 and less than or equal to 0.4 is considered medium-low risk; greater than 0.4 and less than or equal to 0.6 is considered medium risk; greater than 0.6 and less than or equal to 0.8 is considered medium-high risk; and greater than 0.8 is considered high risk. This grading standard is established based on retrospective data of historical structural instability events at industrial sites and combined with laboratory simulation of the accumulation structure evolution curve to ensure that each risk level corresponds to a clear structural state boundary. The training samples of the logistic regression model are derived from pre-collapse data collected from actual accumulation in the past. Fixed weights and bias parameters are obtained through supervised learning. The model stability and discrimination accuracy are confirmed through cross-validation and field testing. In actual operation, a complete logistic regression prediction process is performed on each sample in the updated distribution dataset to obtain the structural instability risk level of all points in the entire grid map. Finally, a risk level result map with spatial distribution significance is generated to guide the design of subsequent adjustment plans for heating power parameters.

[0042] After determining the structural instability risk level, the heating power parameters are automatically adjusted. Initially, the heating power is set to 100% of full power, and then gradually reduced based on the risk level: 80% for high-risk areas, 90% for medium- to high-risk areas, and maintained at the original power level for medium-risk areas and below. The adjusted power parameters are reorganized by grid location, forming an optimized heating distribution plan that defines the power output percentage for each grid cell for the heating device to execute.

[0043] After executing the optimized heating scheme, the sensor network is started to collect the temperature and load-bearing values ​​in the heated state again. The sampling frequency is restored to once every 5 minutes, and the sampling time is 10 minutes to form a verification data set. The temperature fluctuation of each point in the verification data set shall not exceed the range of ±5 degrees Celsius of the temperature value before optimization, and the load-bearing value fluctuation shall not exceed the range of ±20 kPa of the value before optimization. If the above stability judgment conditions are met, it means that the optimized heating scheme is successful and enters the next cycle control process; if not, the failed area is recorded and the risk level judgment step is returned to the risk level judgment step. The heating power of the area is readjusted and iterative optimization is performed until the data fluctuation is within the stable threshold to ensure that both structural safety and thermal efficiency are met.

[0044] A stability assessment mechanism based on sensor data has been designed. Using multi-dimensional data processing and evaluation methods, the system objectively and quantifies temperature and load-bearing data to ensure the feasibility and objectivity of the assessment results. First, after the system implements the optimized heating plan, a sensor network deployed within the carbide slag accumulation area continuously collects real-time temperature and load-bearing data from multiple key areas. The acquisition frequency and period are set by the control platform—for example, once per second for ten minutes—to generate the data sets required for stability verification. After data collection is complete, the system chronologically analyzes temperature and load-bearing changes at each monitoring node. For temperature data, the system determines whether the rate of change is stable throughout the observation period—that is, whether the temperature fluctuates significantly within a unit time. Similarly, for load-bearing data, the system analyzes the magnitude of change within a unit time to determine whether there are significant increases or decreases. Rapid fluctuations in temperature or load within a short period of time indicate thermal or mechanical instability in the area. The system then statistically analyzes the degree of fluctuation in temperature and load-bearing data. Specifically, the system calculates the temperature and pressure variation ranges for each monitoring point throughout the observation period to determine whether they fall within the pre-set stability range. If the data at a particular monitoring point fluctuates minimally over time, it indicates a relatively stable state; otherwise, potential risks are identified. Furthermore, to prevent local anomalies from impacting overall assessments, the system also compares data differences between adjacent monitoring points. If the temperature or load-bearing values ​​at a particular monitoring point differ significantly from those of multiple surrounding points, this indicates a potential concentration of local anomalies and will be marked as unstable. After completing this analysis, the system generates a stability assessment for each monitoring point based on multiple dimensions. These dimensions include temperature change rate, load-bearing change rate, data volatility, and spatial consistency. Each dimension is assigned a set of criteria and stability thresholds. The system then assigns a score based on the assessment results for each dimension, resulting in a comprehensive assessment of the stability of the entire carbide slag accumulation area. If the overall score meets the preset safety standard, the current temperature and load-bearing conditions are stable, and the optimized heating plan is acceptable. Otherwise, the system automatically adjusts the heating power parameters and repeats the data collection and stability verification process until the stability assessment meets the standards. Through the above methods, the system realizes the continuous monitoring, stability judgment and quantitative analysis of the thermal state of the carbide slag accumulation structure, ensuring that the heating process has structural safety while meeting the drying efficiency.

[0045] S4 includes obtaining the temperature control parameters and load-bearing distribution data corresponding to the optimized heating distribution scheme through the sensor network to obtain an initial efficiency data set; using the data standardization method to process the initial efficiency data set to obtain a standardized efficiency data set; classifying and analyzing the standardized efficiency data set through the support vector machine model to obtain drying efficiency data; if the drying efficiency data is lower than the efficiency requirements preset by the production safety standard, re-collecting the temperature control parameters and load-bearing distribution data through the sensor network to obtain an updated efficiency data set; using the logistic regression model to analyze the updated efficiency data set to determine the risk assessment indicators; adjusting the temperature control parameters according to the risk assessment indicators to obtain an optimized adjustment strategy; verifying the optimized adjustment strategy through the sensor network, obtaining a verification data set, and determining whether the drying efficiency data meets the production safety standards.

[0046] In this implementation, first, based on the optimized heating distribution scheme implemented in the previous stage, the latest temperature control parameters and load-bearing distribution data are collected from the temperature sensors and pressure sensors deployed in the calcium carbide slag accumulation area. The sampling frequency is set to once every 5 minutes, covering all heating control units and corresponding structural units in the entire accumulation area. The temperature control parameter is the real-time output power value of each heating unit, in percentage, ranging from 0 to 100; the load-bearing distribution data is the actual load-bearing value corresponding to each grid unit, in kilopascals, ranging from 0 to 500. After the above two types of data are collected, the initial efficiency data set is constructed according to the spatial correspondence. Each sample contains two fields, namely the temperature control power percentage and the load-bearing value of the corresponding point.

[0047] This initial efficiency dataset then undergoes a normalization step. This normalization utilizes a linear normalization method: the temperature-power percentage is divided by 100 to obtain a normalized temperature value between 0 and 1; the load-bearing value is divided by 500 to obtain a normalized load-bearing value between 0 and 1, forming the standardized efficiency dataset. This processing step ensures that data of different dimensions can be simultaneously input into the support vector machine model, avoiding learning bias caused by differences in magnitude.

[0048] The standardized data set was input into a trained support vector machine model for classification analysis. This support vector machine model uses a Gaussian radial basis kernel function to construct a nonlinear interface in high-dimensional space to determine whether each set of input data meets the drying efficiency requirements. The model was trained to determine whether the current temperature control parameters and load-bearing conditions are sufficient to reduce the moisture content of carbide slag from an initial 30% to below 10% within 48 hours. This determination was based on actual drying curves measured under laboratory and industrial production conditions. The model output label was either "meets" or "does not meet" the efficiency requirements.

[0049] A standardized efficiency dataset was prepared as input. This dataset consists of multiple samples, each containing two numerical fields: the standardized temperature control parameter value and the standardized load distribution value. The temperature control parameter value ranges from 0 to 1, corresponding to the original temperature power percentage divided by 100; the load distribution value ranges from 0 to 1, corresponding to the original pressure value divided by 500. Normalization ensures that all input data are dimensional and on a uniform scale, which improves model calculation stability and accuracy. This standardized dataset was then input into a pre-trained support vector machine model. This model uses a Gaussian radial basis function as the kernel function. Its core purpose is to construct a hyperplane in high-dimensional space to classify the data samples into two categories: one indicating that the drying efficiency requirement is met, and the other indicating that the drying efficiency requirement is not met. During the model training phase, a large amount of labeled historical data was used for learning, including labels for drying results under different combinations of temperature power and structural load. The labels were defined as follows: after 48 hours of drying of the carbide slag, if the moisture content of the target area drops below 10%, it is labeled "satisfied"; if it exceeds 10%, it is labeled "unsatisfied." During the model inference phase, each standardized input sample first undergoes an inner product operation with the model parameter vector and then adds the model bias. The result is then mapped to a kernel function to construct a high-dimensional spatial distance metric. The model then determines the class to which the sample belongs based on the distance from the interface. The final output is a binary classification result: if the sample falls into the "satisfied" category, the output label is 1, indicating that the parameter combination is capable of meeting the drying efficiency requirements under the current conditions; if it falls into the "unsatisfied" category, the output label is 0, indicating insufficient efficiency. This process performs classification analysis on all samples in the entire dataset, and all output labels are aggregated to form drying efficiency data. Statistics are collected on these results, for example, the percentage of samples with an output of 1 is determined to be greater than 80%, which serves as the basis for overall judgment. If the percentage meets the criterion, the overall heating strategy meets the drying efficiency target set by production safety standards; if not, the efficiency is considered low, requiring further optimization steps.

[0050] If the model output indicates that the efficiency requirements are not met, a data update process is initiated, and the sensor network is used again to collect temperature control parameters and load distribution data for the same area at a higher frequency. The sampling frequency is increased to every 2 minutes, and the sampling duration is 10 minutes, ensuring that the data contains sufficient temporal variation information. This new data is constructed into an updated efficiency dataset, processed using the same linear normalization method, and then input into the logistic regression model for risk assessment.

[0051] The logistic regression model calculates the weighted sum of standardized temperature and load-bearing values ​​and model parameters. The result is then mapped to a risk probability value between 0 and 1 using a sigmoid function. This value serves as the risk assessment index. This index reflects the potential impact of adjustments to the current heating scheme on the stability of the stacked structure. Based on empirical parameter settings, when this index is less than 0.3, the structure is stable and the current heating power can be increased by 10%; when it is between 0.3 and 0.6, the structure is sensitive and the heating power can be increased by 5%; when it is above 0.6, the structure enters a high-risk state and the heating power remains unchanged to avoid further exacerbation of structural load concentration. First, an update efficiency dataset is constructed based on the latest temperature control parameters and load-bearing distribution data obtained in the aforementioned supplementary sampling step. Each data record contains two fields: the temperature control parameter, ranging from 0 to 100, corresponding to the current output power percentage of each heating unit; and the load-bearing value, in kilopascals, ranging from 0 to 500, corresponding to the current structural bearing pressure in each grid area. The update efficiency dataset was then normalized using a linear normalization method, with the temperature control parameters divided by 100 and the load-bearing values ​​divided by 500, converting them to standardized values ​​between 0 and 1. This generated a standardized update efficiency dataset. The standardized data was then fed into an established logistic regression model for analysis. The logistic regression model is a binary probabilistic prediction model constructed as a weighted linear combiner. The calculation process is as follows: for each input sample, the standardized temperature and load-bearing values ​​are multiplied by two weight parameters obtained through model training. The sum of the results is then added to the bias parameter to produce an intermediate value. This intermediate value is then fed into a sigmoid function, which maps it to a floating-point number between 0 and 1. The output is a risk assessment index. This risk assessment index represents the safety margin that can be adjusted for heating power while maintaining structural stability under the current input parameter combination. Lower values ​​indicate greater structural stability and greater room for power adjustment. Higher values ​​indicate a high risk of damage, requiring strict limits on further increases in temperature and power. Quantitative zoning judgment is made based on the risk assessment index value: if the risk assessment value is less than 0.3, it means that the structural risk is low and the current temperature power can be increased by 10%; if the assessment value is between 0.3 and 0.6, it means that the structural sensitivity is medium and the temperature power is only allowed to increase by 5%; if the assessment value is higher than 0.6, it is judged that the structural risk is high, the power is not allowed to be increased, and the current heating level is maintained.

[0052] Based on the risk assessment indicators output by the logistic regression model, the temperature control parameters at each grid location were adjusted to generate a new temperature control configuration table. The results were then reorganized into an optimized adjustment strategy. This strategy applied control commands to each heating unit to achieve a new power output configuration. A sensor network was simultaneously activated, continuously collecting temperature control parameter and load distribution data every 5 minutes for 24 hours to construct a validation dataset.

[0053] The validation dataset is re-entered into the original support vector machine model, and the same efficiency determination process as previously described is repeated. If the model determines that the efficiency requirements are met, it indicates that the current optimization strategy has met the drying performance target while maintaining structural safety. If it still does not meet the target, the current risk parameters are retained and the logistic regression assessment, power adjustment, and verification process are repeated until the efficiency determination meets production safety standards, ensuring the safety, feasibility, and stability of the final implementation plan in actual industrial operation.

[0054] S5 includes obtaining updated carbide slag accumulation temperature field data and load-bearing distribution field data through a sensor network to obtain an initial field distribution data set; using a data standardization method to process the initial field distribution data set to obtain a standardized field distribution data set; classifying and analyzing the standardized field distribution data set through a random forest model to obtain feature classification results of the temperature field and load-bearing distribution; if the feature classification results show that the temperature field or load-bearing distribution deviates from a preset stability threshold, re-collecting the temperature field data and load-bearing distribution field data through the sensor network to obtain an updated field distribution data set; using a principal component analysis method to perform dimensionality reduction processing on the updated field distribution data set to obtain a reduced dimensionality feature data set; based on the reduced dimensionality feature data set, analyzing the stability of the temperature field and load-bearing distribution through a logistic regression model to obtain a stability evaluation result; if the stability evaluation result meets the production safety standard, adjusting the heating distribution parameters in real time through the sensor network to obtain optimized temperature field data and load-bearing distribution field data.

[0055] In this implementation, a network of temperature and pressure sensors is deployed within the carbide slag accumulation area, with a density of at least one temperature sensor per square meter and one pressure sensor per two square meters, and a sampling period of five minutes. Temperature field data and load-bearing distribution field data are collected for each grid location within the accumulation area at the current moment. The data collected for each grid cell contains two values: a temperature value (in degrees Celsius, ranging from 0 to 300) and a load-bearing value (in kilopascals, ranging from 0 to 500). All data are arranged by spatial location to form an initial field distribution dataset. This dataset is then standardized using linear normalization, which converts each temperature value by 300 and each load-bearing value by 500, converting them into standardized data between 0 and 1 to generate a standardized field distribution dataset. This normalization process aims to unify the numerical range and prevent numerical differences between different physical quantities from affecting the model's discriminant results during subsequent modeling and analysis.

[0056] Subsequently, the standardized field distribution dataset was input into the random forest classification model. The random forest model consists of 100 decision trees. Each decision tree randomly extracts feature combinations from the sample during the training phase to form multiple feature discrimination paths. During the analysis process, each tree independently makes a path judgment on the standardized temperature value and load-bearing value of a grid point, and the output result is "stable" or "unstable". The model ultimately uses a majority vote to determine the judgment result for each grid point. The preset stability thresholds used by the model are: the temperature variation range shall not exceed 5 degrees Celsius, that is, the temperature fluctuation of a single point within the sampling period shall not exceed plus or minus 5 degrees Celsius of the previous sampling value; the load-bearing fluctuation range shall not exceed 20 kPa, that is, the difference between the current load-bearing value of the point and the historical average load-bearing value shall not exceed plus or minus 20 kPa. These two stability thresholds are obtained through statistical analysis of the fluctuation range of 200 stacking stability operation records in the past three years to ensure the reliability and applicability of the judgment results.

[0057] If the random forest model determines that the temperature field or load distribution deviates from the above stability threshold, a high-frequency supplemental sampling mechanism is implemented. The sampling frequency is increased from every 5 minutes to every 2 minutes, and the sampling duration is 10 minutes. The sampling area covers all grid cells determined to be unstable and their eight adjacent grid cells. The sampled data once again constitutes the updated field distribution dataset and is converted to a standardized format using the same normalization method.

[0058] The feature classification results specifically include the following categories: First, in terms of temperature field, the feature classification results include the following: (1) Identification of local high-temperature areas: The system detects that the temperature values ​​in some areas are significantly higher than the surrounding areas or the historical average temperature, and preliminarily classifies them as local overheating risks; (2) Identification of abnormal temperature gradient areas: There are abnormal temperature differences between adjacent areas, such as sudden temperature rises and falls, which may lead to hidden dangers such as uneven thermal expansion and structural deformation; (3) Identification of areas with drastic temperature fluctuations: The temperature in a certain area fluctuates frequently in a short period of time, which may mean unstable heating control or abnormal heat transfer; (4) Assessment of temperature uniformity level: Whether the overall temperature field is evenly distributed. If large-scale thermal unevenness occurs, it may affect the drying efficiency or local structural safety.

[0059] Secondly, in terms of load distribution, the feature classification results mainly include: (1) Identification of local high-pressure areas: The system identifies that some areas have significantly higher load than other areas, indicating the existence of concentrated stress points, and it is necessary to be vigilant about the risk of local settlement or structural damage; (2) Identification of uneven load areas: There are large differences in load within the area, and there are serious problems of uneven stress distribution, which can easily cause local slip or compaction anomalies; (3) Detection of sudden stress changes: The load changes between adjacent areas are drastic, indicating that there may be sudden settlement, overhead, fracture or structural dislocation; (4) Structural stress symmetry analysis: The system conducts a geometric symmetry assessment of the stress conditions of the overall stacking area. If obvious eccentricity or unilateral concentrated stress is found, it indicates that the overall stability of the structure may be reduced.

[0060] In addition, the classification results also include cross-dimensional comprehensive identification, such as: (1) Identification of thermal coupling anomaly areas: areas with both high temperature and high pressure characteristics are marked as high coupling risk areas and require key monitoring; (2) Determination of structural edge risk areas: The temperature and load-bearing characteristics of the edge positions are separately classified to assess whether there are problems with poor edge heat dissipation or weak support; (3) Identification of interference anomalies: If extreme values ​​or data continuity anomalies occur, the classification model will mark them as possible data acquisition failures or external disturbances, prompting the need for resampling.

[0061] The updated standardized dataset is then fed into the principal component analysis processing module to perform data dimensionality reduction. Specifically, the covariance of all sample matrices in the updated dataset is calculated to generate a covariance matrix. Eigenvalue decomposition is then performed on the covariance matrix to extract the principal component vectors. The first few principal components are selected based on a cumulative contribution greater than 95%, typically the first three to five principal components, as feature data to form a reduced-dimensional feature dataset. This dimensionality reduction step aims to reduce model complexity and redundant information while retaining key feature information that effectively expresses temperature and load-bearing trends.

[0062] The reduced-dimensional feature dataset is then input into the logistic regression model for stability probability prediction. The principal component eigenvector of each sample is multiplied by the weight coefficient in the logistic regression model item by item, and the bias term is added to the sum to form a linear combination value. This value is then input into the Sigmoid function and converted into a probability value between 0 and 1, which is called a stability assessment index. This value represents the probability level that the point is in a structurally stable state under the current temperature and load-bearing conditions. The stability threshold is set to 0.7. If the stability index of a grid point is higher than or equal to 0.7, the point is considered to be in a safe and stable state; otherwise, it is considered to be unstable. This threshold is set based on the physical model of the stacking structure and the experience of industrial safety boundaries to ensure that the classification results neither miss high-risk points nor misjudge the normal state.

[0063] After completing the stability assessment for all grid points, if the overall stability meets production safety standards (i.e., all point indicators are no less than 0.7), the control execution process immediately begins. Based on the stability indicators of each region, a differentiated heating parameter adjustment strategy is implemented: if the stability indicators of all grids in a region are above 0.85, the heating power can be increased by 10%; if the stability indicators are between 0.7 and 0.85, the original power is maintained; if individual grids approach the threshold (i.e., between 0.7 and 0.75), the heating power is reduced by 10%. Based on this strategy, a new heating parameter configuration file is generated, and the adjustment instructions are distributed to each heating unit via the network, enabling real-time dynamic adjustment of heating power. The sensor network then continues to monitor the temperature field and load distribution under the new heating state every 5 minutes, constructing the next round of initial field distribution datasets, achieving closed-loop control and real-time response.

[0064] The stability assessment results are a comprehensive assessment of the current thermal stability of the structure, based on a multi-dimensional analysis of the temperature and load distribution data within the carbide slag accumulation area. This assessment aims to identify whether the current heating process will cause structural instability, localized failure, or other safety risks, and whether further adjustments to the heating strategy are needed to ensure continued stable drying.

[0065] After executing the optimized heating plan, the system obtains the latest temperature field distribution data and load-bearing distribution field data of the area through the sensor network deployed in the calcium carbide slag accumulation area. These data reflect the thermal and stress states of different spatial positions, forming a complete, multi-dimensional "field" information map.

[0066] First, a random forest model was used to perform a preliminary classification analysis on the data. This model extracts features from the temperature and load-bearing values ​​at each location and determines whether these points represent risk areas outside the "normal stability range." The output clearly indicates whether the temperature or pressure in a local area is significantly higher or lower than that in other areas, potentially causing localized thermal expansion, structural deflection, or stress concentration.

[0067] If abnormally high temperatures or uneven load distribution are detected in certain areas during this phase, this indicates a potential trend of instability. To further confirm the persistence of this abnormal state, the system will conduct a second round of data collection in these areas, obtaining more intensive or higher-frequency temperature and load data to construct an "updated field distribution dataset." Because the original data is very high-dimensional (i.e., each area may contain multiple time points, multiple data types, and multiple physical indicators), the system uses principal component analysis to reduce the dimensionality of the updated data. This step aims to extract the core factors or features that best represent the overall trend of change, making subsequent judgments more accurate and efficient.

[0068] After dimensionality reduction, the system uses a logistic regression model to comprehensively analyze these core features to determine whether the current temperature and load distribution are showing a "stable trend" or an "instability trend." This determination is based on the following criteria: temporal continuity: whether the temperature and load data have changed significantly over a short period of time; spatial consistency: whether there are significant differences between adjacent points in the same area; trend direction: whether the temperature or load is continuously rising or falling; structural symmetry: whether the forces across the entire platform are balanced; and historical comparability: whether the current data shows a deteriorating trend compared to previous operating conditions.

[0069] Ultimately, the system will output a clear stability assessment result, which can be a qualitative judgment (such as "stable" or "unstable"), a multi-level classification (such as "safe," "metastable," "critical," or "unstable"), or a score or grade value.

[0070] S6 includes obtaining temperature field data and load-bearing distribution field data of carbide slag accumulation through a sensor network to obtain a real-time distribution data set; using a data standardization method to process the real-time distribution data set to obtain a standardized distribution data set; analyzing the standardized distribution data set through a recurrent neural network model to predict the structural stability change trend and obtain a stability trend prediction result; if the stability trend prediction result shows that the local thermal risk is higher than the preset threshold, re-collecting the temperature field data through the sensor network to obtain an updated temperature data set; using a clustering analysis method to group the updated temperature data set to determine the regional distribution where the local overheating risk is higher than the preset threshold and obtain a hot area data set; based on the hot area data set, judging the local overheating path through a preset threshold to obtain a thermal evolution path distribution; optimizing the thermal evolution path distribution by adjusting the heating distribution parameters to obtain optimized temperature field data and load-bearing distribution field data.

[0071] In this embodiment, a temperature and pressure sensor network deployed in the carbide slag accumulation area is first used to collect the temperature data and load-bearing data of each grid in the entire accumulation area point by point with a sampling period of every 5 minutes to construct a real-time distribution data set. The temperature data of each grid point is in degrees Celsius, and the maximum value is set to 300 degrees Celsius; the load-bearing data is in kilopascals, and the maximum value is set to 500 kilopascals. After the collection is completed, each data point in the data set is standardized separately. Specifically, the temperature value is divided by 300 and the load-bearing value is divided by 500, thereby converting all data into floating-point values ​​between 0 and 1 to form a standardized distribution data set. This standardization step ensures the uniformity of the data dimension, which is beneficial to the stability and accuracy of the subsequent model analysis.

[0072] Then, the standardized distribution data set is fed into the pre-trained recurrent neural network model as input. The model adopts a two-layer long short-term memory network structure, and each layer of the model contains 128 neural units, which is used to analyze the time series data of the past 48 hours and predict the structural evolution trend in the next 12 hours. The input data is the temperature and load-bearing data sequence collected every 5 minutes in the first 48 hours, and the output is the temperature change rate and load-bearing growth rate per hour in the future. The set temperature change risk threshold is a temperature rise of more than 5 degrees Celsius per hour, and the load-bearing growth risk threshold is a pressure increase of more than 30 kPa per hour. If the temperature or load-bearing growth rate of any grid point in the prediction result exceeds the above threshold, it is considered that there is a high-risk local thermal evolution trend, and the high-frequency data re-collection process is initiated.

[0073] After the supplementary sampling process is started, temperature data is collected from the above-mentioned high-risk areas and their adjacent 8 grid points at a frequency of once per minute for a duration of 10 minutes. After the supplementary sampling is completed, the data is processed by the same standardization to form an updated temperature data set. Subsequently, the data set is input into the cluster analysis module for group identification. The cluster analysis uses a density-based spatial clustering algorithm to find adjacent points within a radius of 2 meters for each data point, and divides them into independent cluster areas according to the density of the points. The defined local overheating judgment threshold is: if the temperature values ​​of at least 80% of the points in a cluster area are higher than 250 degrees Celsius, and the temperature change trend is in a positive growth direction, then the area is marked as a "local high-temperature area."

[0074] After identifying the high-heat area, the thermal evolution path distribution is further constructed. This process is implemented using the adjacent grid tracing method. Starting from the grid with the highest temperature in the hot area, the eight adjacent grids are checked point by point to see whether the temperature is increasing and the temperature rise rate exceeds 3 degrees Celsius per hour in the time series. If so, the point is included in the thermal evolution path. This method continues to expand outward until the path expansion conditions are no longer met, eventually forming one or more continuous thermal evolution path distribution maps. This map shows the diffusion trend and direction of local heat within the accumulation area and is an important basis for formulating control strategies.

[0075] Based on this thermal evolution path distribution map, heating parameters are optimized and adjusted for the heating units within the path. The adjustment strategy is as follows: If a grid cell is at the start of the thermal evolution path and the current heating power exceeds 80%, the power at that point is reduced to 60%; if it is in the middle of the thermal evolution path and the temperature is still rising, the power is reduced to 50%; if it is at the end of the path, the power remains unchanged; and no adjustments are made to the remaining areas not covered by the path. All adjustment results are organized into an optimized heating parameter configuration table by grid location, which is distributed in real time by the control module to each heating execution unit for implementation.

[0076] After the adjustment is complete, the optimized temperature and load distribution fields are continuously monitored at a sampling frequency of once every five minutes, and the complete data set is updated. The new data is compared with the evolution path distribution to determine whether the temperature control intervention has been effective. If the thermal evolution path is terminated, the heating optimization strategy is effective; otherwise, the next round of adjustment is initiated.

[0077] S7 includes obtaining temperature distribution data of local overheating areas through a sensor network to obtain a real-time temperature data set; based on the real-time temperature data set, using the K-means clustering analysis method to group the temperature distribution, determine the regional distribution where the local temperature is higher than a preset threshold, and obtain an overheating area data set; judging the temperature anomalies in the overheating area data set through a preset threshold, generating an initial heating control instruction sequence, and obtaining a control instruction data set; using a long short-term memory network model to analyze the control instruction data set, predicting the structural stability change trend, and obtaining a stable trend prediction result; if the stable trend prediction result shows that the local thermal risk is higher than the preset threshold, re-collecting the temperature distribution data and the load-bearing distribution data through the sensor network to obtain an updated distribution data set; based on the updated distribution data set, using the support vector machine model to classify the evolution path of the overheating area, and obtain an optimized overheating evolution path distribution; through the optimized overheating evolution path distribution, generating a dynamically adjusted heating control instruction sequence, and determining a real-time monitoring cycle mechanism.

[0078] In this implementation, a temperature sensor network deployed in the carbide slag accumulation area is first used to collect temperature data from all grids within the local overheating risk area at a frequency of once per minute to construct a real-time temperature dataset. Each record in this dataset contains the number of a grid location and its corresponding temperature value, expressed in degrees Celsius, ranging from 0 to 300. After acquisition, the dataset is then subjected to a K-means cluster analysis method to group the temperature distribution. The cluster analysis begins by setting the number of clusters, K, to 5, a value determined empirically based on the temperature gradient stratification characteristics and thermal diffusion range division commonly seen in actual operating conditions. In the initial stage of the K-means algorithm, five temperature values ​​are randomly selected from the dataset as cluster centers. Each data point is then classified based on its Euclidean distance from the cluster center. Once all points have been classified, the center temperature of each category is recalculated, and the classification operation is repeated until the classification of all points remains unchanged, ultimately forming five temperature distribution regions. Groups with temperature cluster centers above 250 degrees Celsius are labeled "overheated regions," and the associated grids are extracted to form the overheated region dataset. The 250-degree Celsius threshold is set based on the temperature safety upper limit measured during the previous carbide slag drying process. This value is the highest temperature at which long-term heating will not cause internal structural damage under the premise of structural stability.

[0079] Subsequently, temperature anomalies within the overheating zone are identified based on two criteria. The first criterion is whether the current grid temperature exceeds 270 degrees Celsius, a safety upper limit determined by statistically analyzing boundary data from thermal instability experiments on structural materials. The second criterion is whether the temperature at that point has risen by more than 10 degrees Celsius in the last five minutes. If both conditions are met, the grid point is marked as a temperature anomaly and an initial heating control instruction is generated for it. The instruction content includes: the grid number, the current heating power value (as a percentage), the recommended control action (such as reducing power by 20%), the instruction timestamp, and so on. The control instructions for all anomaly points are summarized as a control instruction dataset.

[0080] This control instruction dataset is input into a pretrained long short-term memory (LSTM) neural network model to predict structural stability trends. The model consists of a two-layer LSTM network and a fully connected output layer. Inputs include each grid's historical temperature series, heating power adjustment records, and load-bearing change data. The model predicts the probability of structural stability for each grid within the next six hours, outputting a floating-point number between 0 and 1. The stability risk threshold is set at 0.3, optimized based on historical fault prediction accuracy. This value indicates that if the predicted stability probability for a grid is lower than 0.3, the area is at high risk of structural thermal instability. If the model output indicates that the stability probability of a grid is lower than this threshold, a supplementary data collection process is immediately initiated, recollecting the temperature and load-bearing values ​​for the area over the next 10 minutes at a rate of once per minute to generate an updated distribution dataset.

[0081] This updated distribution dataset is then fed into a support vector machine model for evolution path classification analysis. The support vector machine model uses a radial basis kernel function for high-dimensional mapping. The model was previously trained in a supervised learning process using real-world overheating path data. The input for each sample is the temperature value, the load-bearing value, and its rate of change. The output is a classification label indicating whether the point belongs to the thermal evolution path. After processing each data point, the model outputs the classification result. All grid points identified as belonging to the thermal evolution path are then constructed into path segments based on their spatial proximity. These segments are then merged to form an optimized overheating evolution path distribution map.

[0082] Finally, a dynamically adjusted heating control instruction sequence is regenerated based on the optimized path results. The sequence is formulated as follows: if a grid is the starting point of the path, its original heating power is reduced by 30%; if it is located in the path extension area, the power is reduced by 20%; if it is at the end of the path or in an area not included in the path, the current power is maintained unchanged. All instructions are combined to form a new control instruction set, which is distributed to each heating unit for execution in real time by the central control unit. The control is set to a cycle of every 10 minutes. After each round of control execution is completed, new temperature and load-bearing data are collected again, and the judgment and prediction process is restarted to form an adaptive closed-loop control mechanism based on the optimized path, ensuring that local thermal anomalies are identified and intervened in a timely manner, thereby maintaining the thermal stability and safe operation status of the entire stacking structure.

[0083] like Figure 2 As shown, a carbide slag drying temperature control system based on multi-agent collaboration is also provided, which is used to implement the steps of the carbide slag drying temperature control method based on multi-agent collaboration. The system includes a sensor network module, which collects real-time pressure data and temperature data through a sensor network deployed in the carbide slag accumulation area to obtain the current load-bearing distribution map and temperature distribution map of the carbide slag accumulation; an image processing module, which uses a convolutional neural network model to process the image data according to the load-bearing distribution map and the temperature distribution map to determine the structural instability risk level in the carbide slag accumulation; a heating control module, which reduces the temperature input of the corresponding area by adjusting the power parameters of the heating method if the structural instability risk level is higher than the preset threshold, and obtains an optimized heating distribution scheme; and an efficiency control module. The evaluation module obtains the optimized heating distribution plan and uses the support vector machine model to analyze the impact of the plan on the overall drying efficiency to determine whether it meets the efficiency requirements under the production safety standards; the execution control module adjusts the heating method execution plan in real time if the efficiency requirements under the production safety standards are met to obtain the updated carbide slag accumulation temperature field and load-bearing distribution field; the stability prediction module uses a recurrent neural network model to predict the structural stability change trend in the future based on the updated carbide slag accumulation temperature field and load-bearing distribution field, and determines the potential local thermal evolution path; the instruction generation module generates a targeted heating control instruction sequence through the potential local thermal evolution path to obtain a dynamic monitoring cycle mechanism to maintain structural stability.

[0084] In this system implementation method, the sensor network module first performs the real-time data acquisition task, and all temperature and pressure sensors are connected to the central processing unit through wireless or wired communication, and synchronously collect data in a cycle of every 5 minutes. Each sensor corresponds to a physical grid unit in the stacking area, and the system automatically maps the collected data into a two-dimensional temperature distribution map and a load-bearing distribution map. The image processing module receives the above distribution map data and uses a convolutional neural network model to extract image features. It analyzes the areas of local abnormal temperature rise or structural concentration in the image through multi-layer convolution, pooling and fully connected structures, and outputs the risk level, which is a floating point number between 0 and 1, representing the probability of the current structural instability. When the risk level is greater than the threshold value of 0.7 set by the system, the heating control module is started, and the corresponding heating unit is identified according to the high-risk area marked in the image, and its output power is reduced by 20% to 30% by controlling the power parameters of the heating power supply to generate an optimized heating distribution plan.

[0085] The efficiency evaluation module receives this heating distribution plan, combines it with the drying progress data within the current temperature control period and the sampling period, and inputs it into a pre-trained support vector machine model. The model output is judged as "satisfied" or "not meeting the efficiency requirements under production safety standards." The system's efficiency standard is to reduce the moisture content to no more than 10% within 48 hours. If the output is "satisfied," the execution control module will distribute the optimized heating plan to each heating unit. The controller adjusts the temperature output in real time according to the new configuration. At the same time, the sensor network continuously provides feedback on the latest temperature and load-bearing field changes, forming a new closed-loop.

[0086] The updated field distribution is input into the stability prediction module, which uses a recurrent neural network model to perform time-series analysis on multiple consecutive cycles of data. The model structure includes long-short-term memory (LSTM) units, and the output is the structural stability probability value for each area over the next six hours. If the future temperature rise rate in a local area of ​​the predicted path exceeds 5 degrees Celsius per hour or the load-bearing fluctuation exceeds 30 kilopascals, the system identifies it as a potential thermal evolution path. Based on this path structure and the system's heating unit locations and adjustment capabilities, the instruction generation module then generates a sequence of heating control instructions. Each instruction includes the unit location, current state, recommended control range, and execution time period. All instructions are sent to the execution module via a control channel to complete dynamic control. The control system updates the execution strategy every 10 minutes, implementing a closed-loop "perception-recognition-prediction-adjustment" cycle to ensure the dynamic coordinated operation of drying efficiency and stacking structure safety.

[0087] In the present embodiment, a "local overheating area" refers to a specific region in the temperature field data collected by the sensor network, identified by cluster analysis methods (such as K-means) as having a temperature higher than the overall temperature distribution, representing an abnormal temperature distribution under the current state. A "local thermal risk exceeding a preset threshold" is a risk assessment result derived from a trend analysis of the stability of the carbide slag accumulation structure over a future time period using a predictive model (such as a recurrent neural network or support vector machine) based on the aforementioned temperature and load-bearing data. This indicates the potential for structural instability or thermal evolution in that region. Therefore, the two are technically different. While "local overheating area" focuses on identifying anomalies in the current spatial state, "local thermal risk exceeding a preset threshold" dynamically predicts the future evolution trend of the abnormal region. While the two can form a causal relationship, they are not interchangeable. In implementation, the system typically first identifies the local overheating area and then assesses its thermal risk level to determine whether to optimize heating control instructions and dynamically monitor and adjust them.

[0088] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A carbide slag drying temperature control method based on multi-agent collaboration, characterized in that: The method comprises: S1. Collect real-time pressure data and temperature data through a sensor network deployed in the carbide slag accumulation area to obtain the current load distribution map and temperature distribution map of the carbide slag accumulation; S2. Based on the load-bearing distribution map and the temperature distribution map, a convolutional neural network model is used to process the map data to determine the structural instability risk level of the carbide slag accumulation; S3. If the structural instability risk level is higher than a preset threshold, the temperature input of the corresponding area is reduced by adjusting the power parameters of the heating method to obtain an optimized heating distribution plan; S4. After obtaining the optimized heating distribution scheme, use the support vector machine model to analyze the impact of the scheme on the overall drying efficiency to determine whether it meets the efficiency requirements under the production safety standards; S5. If the efficiency requirements under the production safety standards are met, the heating method execution plan is adjusted in real time to obtain an updated carbide slag accumulation temperature field and load distribution field; S6. Based on the updated carbide slag accumulation temperature field and load-bearing distribution field, a recurrent neural network model is used to predict the structural stability change trend in the future and determine the potential local thermal evolution path; S7. Generate a targeted heating control instruction sequence through the potential local thermal evolution path to obtain a dynamic monitoring cycle mechanism to maintain structural stability.

2. The method for controlling temperature of carbide slag drying based on multi-agent collaboration according to claim 1, wherein: Said S1 comprises: A sensor network is deployed in a grid-like manner in the carbide slag accumulation area to obtain pressure and temperature data and generate an initial data set. Based on the initial data set, the data cleaning method is used to remove noise data to obtain cleaned pressure data and temperature data; The load-bearing distribution is calculated by using the interpolation algorithm through the pressure data after cleaning, and the load-bearing distribution diagram of the calcium carbide slag accumulation is generated; The temperature distribution is calculated by using the interpolation algorithm through the cleaned temperature data to generate the temperature distribution map of the carbide slag accumulation; If there is an area in the load-bearing distribution map where the abnormal pressure level is higher than a preset threshold, the sensor network is used to recollect the area data to obtain updated pressure data; If there is an area in the temperature distribution map where the temperature anomaly is higher than a preset threshold, the sensor network is used to recollect the area data to obtain updated temperature data; Based on the updated pressure data and temperature data, a weighted average algorithm is used to fuse multiple sets of data to generate optimized load-bearing distribution maps and temperature distribution maps.

3. The method for temperature control of carbide slag drying based on multi-agent collaboration according to claim 1, characterized in that: The S2 includes: The load-bearing distribution map and temperature distribution map of the carbide slag accumulation area are obtained through the sensor network to generate initial image data; The initial image data is standardized by using a data preprocessing method to obtain standardized image data; The standardized image data is subjected to feature extraction through a convolutional neural network to obtain a feature vector; Based on the feature vector, a convolutional neural network is used to calculate the probability of occurrence of local overheating areas and obtain the probability distribution; If there is an area in the probability distribution that is higher than the preset threshold, the load-bearing and temperature data of the area are recollected through the sensor network to obtain updated image data; The updated image data is analyzed through a convolutional neural network to calculate the structural instability risk level and obtain the risk assessment result; According to the risk assessment results, the area division method was used to grade and mark the carbide slag accumulation area, and a graded distribution map was obtained.

4. The method for temperature control of carbide slag drying based on multi-agent collaboration according to claim 1, wherein: The S3 includes: The temperature and load-bearing data of the carbide slag accumulation area are obtained through the sensor network to generate an initial distribution data set; The data normalization method is used to process the initial distribution data set to obtain a standardized data set; The standardized data set is subjected to feature extraction through a convolutional neural network to obtain a feature vector; If there is a risk indicator in the feature vector that is higher than the preset threshold, the temperature data and load-bearing data of the corresponding area are recollected through the sensor network to generate an updated distribution data set; A logistic regression model is used to analyze the update distribution data set to determine the structural instability risk level; Adjust heating power parameters according to the structural instability risk level and generate an optimized heating distribution plan; The optimized heating distribution scheme is verified through the sensor network, and a verification data set is obtained to determine the stability of the temperature and load-bearing data.

5. The method for temperature control of carbide slag drying based on multi-agent collaboration according to claim 1, characterized in that: The S4 includes: The temperature control parameters and load-bearing distribution data corresponding to the optimized heating distribution scheme are obtained through the sensor network to obtain the initial efficiency data set; The initial efficiency data set is processed using a data standardization method to obtain a standardized efficiency data set; The standardized efficiency data set was classified and analyzed using the support vector machine model to obtain the drying efficiency data; If the drying efficiency data is lower than the efficiency requirement preset by the production safety standard, the temperature control parameters and load distribution data are re-collected through the sensor network to obtain an updated efficiency data set; A logistic regression model was used to analyze the update efficiency dataset to determine the risk assessment indicators; Adjust the temperature control parameters according to the risk assessment indicators to obtain the optimized adjustment strategy; The optimization and adjustment strategy is verified through the sensor network, and the verification data set is obtained to determine whether the drying efficiency data meets the production safety standards.

6. The method for temperature control of carbide slag drying based on multi-agent collaboration according to claim 1, characterized in that: The S5 includes: Acquire updated carbide slag accumulation temperature field data and load-bearing distribution field data through the sensor network to obtain the initial field distribution data set; The initial field distribution data set is processed by a data standardization method to obtain a standardized field distribution data set; The standardized field distribution data set was classified and analyzed using the random forest model to obtain the characteristic classification results of the temperature field and load-bearing distribution; If the feature classification results show that the temperature field or load-bearing distribution deviates from the preset stability threshold, the temperature field data and load-bearing distribution field data are recollected through the sensor network to obtain an updated field distribution data set; The principal component analysis method is used to reduce the dimension of the update field distribution data set to obtain the reduced dimension feature data set; Based on the dimensionality-reduced feature dataset, the stability of the temperature field and load-bearing distribution is analyzed using a logistic regression model to obtain the stability assessment results. If the stability assessment results meet the production safety standards, the heating distribution parameters are adjusted in real time through the sensor network to obtain optimized temperature field data and load-bearing distribution field data.

7. The method for temperature control of carbide slag drying based on multi-agent collaboration according to claim 1, characterized in that: The S6 includes: The temperature field data and load distribution field data of carbide slag accumulation are acquired through the sensor network to obtain a real-time distribution data set; The real-time distributed data set is processed using a data standardization method to obtain a standardized distributed data set; By analyzing the standardized distribution data set through the recurrent neural network model, the structural stability change trend is predicted and the stability trend prediction result is obtained; If the stable trend prediction result shows that the local thermal risk is higher than the preset threshold, the temperature field data is recollected through the sensor network to obtain an updated temperature dataset; Cluster analysis is used to group the updated temperature dataset to determine the distribution of areas where the local overheating risk is higher than the preset threshold, thus obtaining the hot area dataset. Based on the hot area data set, the local overheating path is determined by the preset threshold and the thermal evolution path distribution is obtained; By adjusting the heating distribution parameters to optimize the thermal evolution path distribution, the optimized temperature field data and load-bearing distribution field data are obtained.

8. The method for temperature control of carbide slag drying based on multi-agent collaboration according to claim 1, characterized in that: The S7 includes: The temperature distribution data of the local overheating area is obtained through the sensor network to obtain a real-time temperature data set; Based on the real-time temperature data set, the K-means clustering analysis method is used to group the temperature distribution, determine the distribution of areas where the local temperature is higher than the preset threshold, and obtain the overheating area data set; The temperature anomaly points in the overheating area data set are judged by a preset threshold value, and an initial heating control instruction sequence is generated to obtain a control instruction data set; The long short-term memory network model is used to analyze the control instruction data set, predict the structural stability change trend, and obtain the stable trend prediction result.

9. The method for controlling temperature of carbide slag drying based on multi-agent collaboration according to claim 8, characterized in that: The S7 further includes: If the stable trend prediction result shows that the local thermal risk is higher than the preset threshold, the temperature distribution data and load-bearing distribution data are recollected through the sensor network to obtain an updated distribution data set; Based on the updated distribution data set, the support vector machine model is used to classify the evolution path of the overheating area and obtain the optimized overheating evolution path distribution; Through the optimized distribution of overheating evolution paths, a dynamically adjusted heating control instruction sequence is generated to determine the real-time monitoring cycle mechanism.

10. A carbide slag drying temperature control system based on multi-agent collaboration, used to implement the steps of the carbide slag drying temperature control method based on multi-agent collaboration according to any one of claims 1 to 9, characterized in that: The system comprises: The sensor network module collects real-time pressure and temperature data through the sensor network deployed in the carbide slag accumulation area to obtain the current load distribution map and temperature distribution map of the carbide slag accumulation; The image processing module uses a convolutional neural network model to process the image data based on the load distribution map and temperature distribution map to determine the structural instability risk level in the calcium carbide slag accumulation; The heating control module adjusts the power parameters of the heating method to reduce the temperature input of the corresponding area if the structural instability risk level is higher than the preset threshold, thereby obtaining an optimized heating distribution plan; The efficiency evaluation module obtains the optimized heating distribution plan and uses a support vector machine model to analyze the impact of the plan on the overall drying efficiency to determine whether it meets the efficiency requirements under production safety standards; The execution control module, if the efficiency requirements under the production safety standards are met, adjusts the heating method execution plan in real time to obtain the updated carbide slag accumulation temperature field and load distribution field; The stability prediction module uses a recurrent neural network model to predict the structural stability change trend in the future based on the updated carbide slag accumulation temperature field and load-bearing distribution field, and determines the potential local thermal evolution path; The instruction generation module generates a targeted heating control instruction sequence through the potential local thermal evolution path, and obtains a dynamic monitoring cycle mechanism to maintain structural stability.

Images