A medium-frequency electric furnace operation monitoring platform

Abstract

This invention relates to the field of medium-frequency induction furnace monitoring technology and discloses a medium-frequency induction furnace operation monitoring platform. The system's temperature field modeling module generates multi-dimensional temperature distribution codes based on real-time thermal imaging images; the operation trajectory prediction module predicts the thermal field evolution trajectory based on the initial temperature code and the initial thermal imaging image; the power allocation strategy module generates a power control parameter set by combining the current temperature code and the predicted trajectory; the cooling dynamic adaptation module analyzes the flow characteristics of the cooling medium based on the power control parameter set and generates cooling path control results; the abnormal state intervention module extracts temperature fluctuation characteristics and thermal field offsets to generate an abnormal intervention parameter set; and the operation performance evaluation module analyzes the target temperature zone distribution and stabilization time to generate an operation performance data table. Through multi-module collaboration, this platform achieves comprehensive monitoring, dynamic control, and performance evaluation of the medium-frequency induction furnace operation process, improving the accuracy and adaptability of operation monitoring.

Description

Technical Field

[0001] This invention relates to the field of medium-frequency electric furnace monitoring technology, specifically to a medium-frequency electric furnace operation monitoring platform. Background Technology

[0002] Medium-frequency induction furnaces are important heating equipment in industrial production, widely used in metallurgy, casting, heat treatment, and other fields. During their operation, the stability of the temperature field, the rationality of power distribution, and the adaptability of the cooling system directly affect product quality and equipment lifespan. However, existing medium-frequency induction furnace operation monitoring technologies have many limitations.

[0003] Traditional temperature monitoring relies heavily on single-point temperature sensors, which can only acquire localized temperature data and cannot comprehensively reflect the complex temperature field distribution within the furnace. This limitation makes it difficult for operators to accurately determine the heating status of the materials inside the furnace, easily leading to problems such as localized overheating or uneven heating. Furthermore, current technologies lack effective prediction of the evolution of the thermal field, often only allowing for delayed adjustments based on real-time temperature data, making it difficult to adapt to the dynamically changing operating conditions of medium-frequency induction furnaces.

[0004] In terms of power allocation, existing control strategies are mostly based on setting fixed parameters based on experience, failing to dynamically adjust according to real-time temperature field characteristics and thermal field evolution trends. This results in a mismatch between power output and actual heating demand, not only causing energy waste but also potentially affecting heating efficiency and material properties due to excessively high or low power.

[0005] As a crucial component for ensuring the safe operation of medium-frequency electric furnaces, the existing cooling system often employs a fixed cooling path design. This design fails to adjust the flow path and flow rate of the cooling medium in real time based on changes in the furnace surface temperature distribution and thermal field. When the furnace temperature rises abnormally in certain areas, the fixed cooling path cannot achieve precise cooling, potentially leading to overheating damage, increased maintenance costs, and downtime risks.

[0006] Furthermore, existing anomaly identification methods largely rely on simple temperature threshold judgments, lacking in-depth analysis of temperature fluctuation characteristics and thermal field shift patterns, which easily leads to false alarms or missed alarms. Operational performance evaluations are also mostly limited to basic data statistics such as energy consumption and heating time, failing to incorporate key indicators such as temperature field stability and thermal field uniformity for comprehensive evaluation, thus failing to fully reflect the actual operating status of the equipment. These problems collectively result in incomplete monitoring, inaccurate control, and untimely response during the operation of medium-frequency electric furnaces, hindering the improvement of production efficiency and the guarantee of equipment safety. Summary of the Invention

[0007] The purpose of this invention is to provide a medium-frequency electric furnace operation monitoring platform to solve the problems mentioned in the background art.

[0008] To achieve the above objectives, the present invention provides a medium-frequency electric furnace operation monitoring platform, the platform comprising:

[0009] The temperature field modeling module is used to generate multi-dimensional temperature distribution codes based on the real-time thermal imaging of the medium-frequency electric furnace.

[0010] The trajectory prediction module predicts the thermal field evolution trajectory during the operation of the electric furnace based on the initial temperature code in the multi-dimensional temperature distribution code and the initial thermal imaging map.

[0011] The power allocation strategy module generates a power control parameter set based on the current temperature code in the multi-dimensional temperature distribution code and the predicted thermal field evolution trajectory.

[0012] The cooling dynamic adaptation module analyzes the flow characteristics of the cooling medium on the furnace surface based on the power regulation parameter set and generates cooling path regulation results.

[0013] The abnormal state intervention module extracts temperature fluctuation characteristics and thermal field offset based on the cooling path control results, and generates an abnormal intervention parameter set.

[0014] The operational performance evaluation module analyzes the target temperature zone distribution and stabilization duration based on the abnormal intervention parameter set, and generates an electric furnace operational performance data table.

[0015] Preferably, the temperature field modeling module consists of a feature extraction unit and a field reconstruction unit. The feature extraction unit decomposes the thermal image into discrete temperature feature vectors, and the field reconstruction unit converts the discrete temperature feature vectors into a temperature field distribution map to generate a multi-dimensional temperature distribution code.

[0016] Preferably, the temperature field modeling module performs a training process before generating a multi-dimensional temperature distribution code: acquiring historical thermal images and inputting them into the feature extraction unit to generate a temperature feature vector;

[0017] The temperature feature vector is reconstructed into a temperature field distribution map through the field reconstruction unit;

[0018] The differences between the reconstructed map and the actual thermal image are compared and the network parameters are updated until convergence.

[0019] Preferably, the trajectory prediction module superimposes a disturbance factor on the initial temperature code, and learns to remove the disturbance using the initial thermal imaging image and corresponding smelting process parameters as conditions, thereby predicting the thermal field evolution trajectory during the operation of the electric furnace.

[0020] Preferably, before predicting the thermal field evolution trajectory, the trajectory prediction module optimizes the network parameters by maximizing the probability distribution similarity of the thermal field evolution trajectory.

[0021] Preferably, the power allocation strategy module includes: a thermal field equilibrium assessment unit, which calculates the regional temperature difference index based on the current temperature code and the thermal field evolution trajectory;

[0022] The power timing allocation unit calculates the power stage allocation index based on the current temperature code and the thermal field evolution trajectory.

[0023] The efficiency optimization unit generates energy efficiency improvement indicators based on the current temperature encoding.

[0024] The power parameter synthesis unit integrates the regional temperature difference index, power stage allocation index, and energy efficiency improvement index to generate a power regulation parameter set.

[0025] Preferably, the cooling dynamic adaptation module executes the following process: extracting the adsorption characteristics of the medium on the furnace surface based on the power control parameter set; generating an adsorption-diffusion parameter set by combining environmental variables to predict the medium diffusion characteristics; and generating a cooling path control result by adjusting the medium flow rate distribution ratio according to the adsorption-diffusion parameter set.

[0026] Preferably, the abnormal state intervention module executes the following process: monitoring the temperature fluctuation rate and thermal field offset based on the cooling path control results;

[0027] Analyze the impact of fluctuation range on metallurgical reaction rate to generate reaction rate intervention data;

[0028] Based on the reaction rate intervention data, the medium distribution path in the target temperature zone is dynamically adjusted to generate an abnormal intervention parameter set.

[0029] Preferably, the operational performance evaluation module executes the following process: collecting target temperature zone distribution data for different time periods based on the abnormal intervention parameter set;

[0030] Analyze the temperature range stability duration and energy consumption trends to generate target temperature range performance parameters;

[0031] Based on the performance parameters of the target temperature zone, adjust the temperature control range to generate an electric furnace operation performance data table.

[0032] Preferably, the platform further includes a dynamic verification module: when the feature value of a specific region in the multi-dimensional temperature distribution code generated by the temperature field modeling module exceeds a threshold, the corresponding sub-region map is extracted from the thermal imaging map;

[0033] The sub-region map is input into the abnormal state intervention module for secondary analysis;

[0034] The abnormal intervention parameter set is updated based on the results of the secondary analysis.

[0035] Compared with the prior art, the beneficial effects of the present invention are:

[0036] This medium-frequency electric furnace operation monitoring platform achieves comprehensive and dynamic monitoring and control of the furnace operation process through the collaborative work of multiple modules. The temperature field modeling module generates multi-dimensional temperature distribution codes based on real-time thermal imaging images, breaking through the limitations of traditional single-point monitoring. It can capture the spatial distribution characteristics of the temperature field inside the furnace more comprehensively and meticulously, enabling operators to clearly grasp the temperature status of each area inside the furnace and providing rich basic data for control.

[0037] The trajectory prediction module combines initial temperature coding and initial thermal imaging to predict the evolution trajectory of the thermal field, changing the previous mode of relying solely on real-time data for lagging adjustments. By predicting the trend of thermal field changes in advance, valuable adjustment time is gained for power regulation and cooling adaptation, making the entire monitoring and control process more forward-looking and better able to adapt to the dynamically changing operating environment of the electric furnace.

[0038] The power allocation strategy module generates a power control parameter set based on the current temperature code in the multi-dimensional temperature distribution code and the predicted thermal field evolution trajectory, breaking away from the traditional fixed parameter control mode that relies on experience. This dynamic control method based on real-time status and future trends enables the power output to better match the actual heating demand in the furnace, avoiding power waste or insufficiency, and making the power allocation more in line with the operating conditions.

[0039] The dynamic cooling adaptation module analyzes the flow characteristics of the cooling medium on the furnace surface based on a set of power control parameters and generates cooling path adjustment results, breaking the limitations of fixed cooling paths. By adjusting the flow path of the cooling medium according to the real-time thermal state of the furnace, targeted cooling can be achieved, improving the cooling system's adaptability to changes in furnace thermal conditions and reducing the risk of equipment damage due to improper cooling.

[0040] The abnormal state intervention module generates an abnormal intervention parameter set by extracting temperature fluctuation characteristics and thermal field offsets from the cooling path control results. Compared with traditional simple threshold judgment, it can more accurately identify abnormal states. This anomaly identification method based on multi-feature analysis can promptly detect potential operational risks, provide precise intervention basis for anomaly handling, and enhance the stability of equipment operation.

[0041] The operational performance evaluation module analyzes the target temperature zone distribution and stabilization duration based on anomaly intervention parameter sets, generating an electric furnace operational performance data table, breaking through the limitations of traditional basic data statistics. By comprehensively evaluating the equipment's operating status through key indicators, operators can gain a comprehensive understanding of the electric furnace's performance, providing comprehensive reference information for optimizing operational strategies. The various modules work synergistically, forming a complete closed loop from temperature sensing, trend prediction, power control, cooling adaptation, anomaly intervention to performance evaluation, improving the overall integrity and accuracy of medium-frequency electric furnace operation monitoring. Attached Figure Description

[0042] Figure 1 This is a schematic diagram of the working principle of the medium-frequency electric furnace operation monitoring platform described in this invention;

[0043] Figure 2 Flowchart for training modules for temperature field modeling;

[0044] Figure 3 Here is a basic flowchart of the trajectory prediction module;

[0045] Figure 4 This is a flowchart illustrating the components of the power allocation strategy module. Detailed Implementation

[0046] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0047] Please see Figure 1 The present invention provides a medium-frequency electric furnace operation monitoring platform, the platform comprising: a temperature field modeling module, an operation trajectory prediction module, a power distribution strategy module, a cooling dynamic adaptation module, an abnormal state intervention module, and an operation performance evaluation module.

[0048] The temperature field modeling module generates a multi-dimensional temperature distribution code from real-time thermal imaging, which includes the spatial distribution and temporal variation characteristics of the temperature field. The operation trajectory prediction module, based on the initial temperature code and initial thermal imaging, combined with smelting process parameters, predicts the thermal field evolution trajectory during electric furnace operation. The power allocation strategy module calculates regional temperature differences, power allocation stages, and energy efficiency optimization indicators based on the current temperature code and the predicted thermal field evolution trajectory, generating a power control parameter set. The cooling dynamic adaptation module analyzes the medium flow characteristics in the power control parameter set and generates cooling path control results to optimize cooling efficiency. The abnormal state intervention module extracts temperature fluctuation characteristics and thermal field offsets, generating an abnormal intervention parameter set to dynamically adjust the temperature zone distribution. The operation performance evaluation module integrates the abnormal intervention parameter set and target temperature zone distribution data to generate an electric furnace operation performance data table for evaluating system stability and energy consumption performance.

[0049] Example 1: See Figure 2 The temperature field modeling module is a core component of the medium-frequency electric furnace operation monitoring platform, responsible for converting real-time acquired thermal imaging data into multi-dimensional temperature distribution codes that can be used for analysis. This module consists of a feature extraction unit and a field reconstruction unit, which respectively complete the feature decomposition and temperature field reconstruction tasks of the thermal imaging data.

[0050] The feature extraction unit employs a deep convolutional neural network architecture to perform hierarchical feature extraction on the input thermal image. The thermal image is first preprocessed, including noise suppression, edge enhancement, and normalization operations to improve the accuracy of feature extraction. The convolutional neural network progressively extracts the temperature distribution features from the thermal image through multiple layers of convolution and pooling operations. The initial convolutional layer identifies the basic temperature gradient distribution, the intermediate layers capture the spatial correlation between local hotspots and cold areas, and the deep network extracts the macroscopic trend of the global temperature field. Finally, the feature extraction unit outputs discrete temperature feature vectors, each vector corresponding to the temperature statistics of a specific region in the thermal image, including peak temperature, average temperature, temperature change rate, and spatial distribution pattern.

[0051] The field reconstruction unit receives discrete temperature feature vectors from the feature extraction unit and reconstructs a continuous temperature field distribution map through a fully connected network and deconvolution operations. The fully connected network maps the discrete feature vectors to a high-dimensional space, generating a latent representation of the temperature field. The deconvolution network then progressively upsamples this latent representation to recover the spatial details of the temperature field. The output of the field reconstruction unit is a multi-dimensional temperature distribution code, which not only contains spatial distribution information of temperature but also covers the temporal trend of temperature changes, such as heat conduction rate, regional temperature difference evolution, and hotspot migration paths.

[0052] Before the temperature field modeling module is put into use, a training process is required to optimize the network parameters. The training data consists of historical thermal images, covering temperature distribution under different smelting stages and operating conditions. The feature extraction unit processes each thermal image to generate a corresponding temperature feature vector. The field reconstruction unit receives these vectors and attempts to reconstruct the original temperature field distribution map. During training, the difference between the reconstructed map and the actual thermal image is calculated, and the backpropagation algorithm is used to adjust the network weights to gradually reduce the reconstruction error. The training objectives include improving the accuracy of temperature gradient reconstruction, reducing local temperature distribution distortion, and enhancing robustness to noise.

[0053] After training, the temperature field modeling module can process thermal imaging data collected during the operation of the electric furnace in real time. The input thermal imaging image, after feature extraction and field reconstruction, generates a multi-dimensional temperature distribution code that can be used for analysis by the module. This code not only reflects the static distribution of the current temperature field but also implies the dynamic trend of temperature changes, providing a data foundation for trajectory prediction and power allocation.

[0054] The feature extraction unit was designed with the characteristics of electric furnace thermal imaging data in mind. Since the temperature distribution in medium-frequency electric furnaces is typically non-uniform, with potential localized high-temperature zones or rapidly changing temperature gradients, the deep structure of convolutional neural networks can effectively capture these complex patterns. Simultaneously, the network employs residual connections to mitigate the gradient vanishing problem, ensuring the effective transfer of deep features.

[0055] The deconvolutional network structure of the field reconstruction unit is optimized to balance computational efficiency and reconstruction accuracy. The deconvolutional layers employ a progressive upsampling strategy to gradually recover detailed information of the temperature field. During training, a perceptual loss function is introduced to ensure that the reconstructed temperature field not only approximates the real data at the pixel level but also maintains consistency across higher-order features.

[0056] The training data for the temperature field modeling module needs to cover a wide range of operating conditions, including thermal images under different power settings, cooling conditions, and smelting materials. Data augmentation techniques such as rotation, translation, and noise addition are used to improve the model's generalization ability. After training, the module can adapt to temperature fluctuations and noise interference in actual operation and stably output high-precision temperature distribution codes.

[0057] The output of this module is not only used for real-time monitoring but also provides crucial input for the prediction and optimization modules. The multi-dimensional temperature distribution coding structure is designed with compatibility in mind for downstream modules, ensuring efficient data transmission and processing. For example, the regional temperature difference information contained in the coding can be directly used for the equilibrium evaluation of the power allocation strategy module, while temperature change trend data supports the long-term evolution analysis of the trajectory prediction module.

[0058] In actual deployment, the temperature field modeling module employs a distributed computing architecture to improve processing speed. The tasks of acquiring thermal imaging data, extracting features, and reconstructing the field are distributed across different computing nodes to ensure real-time requirements are met. The module also possesses dynamic adjustment capabilities; when abnormal temperature distributions are detected, a higher-precision local reconstruction mode can be triggered to provide more detailed temperature field analysis.

[0059] Through the above implementation, the temperature field modeling module can efficiently and accurately extract key temperature information from the thermal imaging data of the medium-frequency furnace, providing reliable data support for the operation of the entire monitoring platform. The module's design fully considers the complexity and real-time requirements of actual industrial environments, ensuring stable operation under various working conditions.

[0060] Example 2: See Figure 3 The operation trajectory prediction module is a crucial component of the medium-frequency electric furnace operation monitoring platform. It is responsible for predicting the thermal field evolution trend over a future period based on the current temperature distribution. This module employs deep learning to establish a thermal field evolution model, analyzing historical and real-time monitoring data to predict temperature field changes. The module's inputs include the initial temperature code from the multi-dimensional temperature distribution code generated by the temperature field modeling module and the initial thermal imaging image. The output is the predicted thermal field evolution trajectory during the electric furnace's operation.

[0061] The core of this module is a prediction model based on a Conditional Generative Adversarial Network (CGAN). The generator network receives the initial temperature code as input, and the initial thermal image and smelting process parameters as conditional inputs. To enhance the model's robustness, a Gaussian-distributed random perturbation is superimposed on the initial temperature code to simulate measurement noise and uncertainties that may exist in actual operation. The generator network learns stepwise through a multi-layer neural network structure to remove these perturbations and predicts the thermal field state at several future time steps. The discriminator network is responsible for evaluating the similarity between the predicted trajectory output by the generator and the actual observed trajectory, guiding the generator's training process.

[0062] During the model training phase, a training sample set is constructed using a large amount of historical operational data. Each sample contains initial temperature field data, process parameters, and actual temperature field changes over multiple time steps. During training, the generator continuously adjusts its parameters to produce predictions that more closely approximate the actual thermal field evolution, while the discriminator continuously improves its ability to distinguish between real and generated data. Through this adversarial training mechanism, the model gradually learns the physical laws and statistical characteristics of thermal field changes. After training, the model can predict future thermal field evolution trends based on the current temperature field state and process conditions.

[0063] During the prediction process, the module first receives the initial temperature code and initial thermal image provided by the temperature field modeling module. This data, after preprocessing, is input into the trained prediction model. The model outputs temperature field predictions for multiple future time steps, forming a complete thermal field evolution trajectory. This trajectory not only includes the spatial distribution changes of the temperature field but also reflects key information such as hotspot migration paths and regional temperature difference trends. The temporal resolution of the prediction results can be adjusted according to actual needs to adapt to monitoring and control requirements at different time scales.

[0064] To improve prediction accuracy, the module employs a multi-scale prediction strategy. For short-term predictions, the model focuses on capturing rapid changes in the temperature field; for medium- and long-term predictions, the model pays more attention to the overall evolution trend of the temperature field. This multi-scale approach balances prediction accuracy and time span, meeting the needs of different application scenarios. The prediction results also include uncertainty estimates, reflecting the reliability of the predictions.

[0065] The output of the trajectory prediction module is used by multiple downstream modules, such as the power allocation strategy module. The predicted thermal field evolution trajectory can help identify potential temperature anomalies in advance, providing forward-looking guidance for power adjustment and cooling strategy optimization. The module also has online update capabilities, which can continuously revise the prediction model based on the latest monitoring data, improving the accuracy of predicting future thermal field changes.

[0066] In practical applications, this module needs to handle thermal field prediction problems under various complex operating conditions. Different smelting stages, different material properties, and different power settings all affect the evolution of the thermal field. Therefore, the model design fully considers the influence of these factors, and by introducing process parameters as input conditions, the prediction results can adapt to different operating conditions. The module also has the ability to handle sudden changes in operating conditions. When a drastic change in the temperature field is detected, it can quickly adjust the prediction strategy and provide timely prediction updates.

[0067] The module employs a distributed computing architecture to meet the computational demands of real-time prediction. Prediction tasks are distributed across multiple computing nodes for parallel execution, ensuring that prediction calculations are completed within strict time constraints. The module also incorporates fault-tolerance mechanisms to maintain basic prediction functionality even when some computing nodes fail. These design features guarantee reliable operation of the module in industrial environments.

[0068] The performance of the trajectory prediction module depends on the quality and quantity of training data. Before model deployment, historical operational data covering various typical operating conditions needs to be collected, including temperature field change records under normal and abnormal operating conditions. The data preprocessing stage requires cleaning, aligning, and normalizing the raw data to eliminate the influence of measurement errors and noise. Model training adopts a progressive strategy, first pre-training on a small dataset and then gradually expanding the training data scale to improve training efficiency and model performance.

[0069] The module also incorporates various verification mechanisms to assess the reliability of the prediction results. These include residual analysis, consistency checks, and physical law compliance checks. These mechanisms allow for real-time monitoring of prediction quality during the prediction process, promptly identifying and correcting potential prediction biases. When the prediction results show a significant discrepancy with physical laws, the module will automatically trigger a re-prediction or issue an early warning signal.

[0070] The trajectory prediction module, working collaboratively with other modules on the platform, enables comprehensive monitoring and control of the medium-frequency induction furnace's operating status. By accurately predicting the evolution trend of the thermal field, this module provides crucial information for optimizing power distribution, adjusting cooling strategies, and preventing abnormal operating conditions. The module's predictive capabilities continuously improve with the accumulation of operating data, forming a virtuous cycle of learning and continuously enhancing the performance of the entire monitoring platform.

[0071] Example 3: See Figure 4 The power allocation strategy module and the cooling dynamic adaptation module together constitute the intelligent control system of the medium-frequency electric furnace operation monitoring platform. Based on temperature field data and predicted trajectories, these two modules work together to optimize power allocation and dynamically adjust cooling parameters, achieving refined control of the electric furnace operation process.

[0072] The power allocation strategy module consists of four functional units: a thermal field equilibrium assessment unit, a power timing allocation unit, an efficiency optimization unit, and a power parameter synthesis unit. The thermal field equilibrium assessment unit receives the current temperature code from the temperature field modeling module and the thermal field evolution trajectory data from the operating trajectory prediction module. By analyzing the temperature differences in different regions, it calculates a regional temperature difference index. This index reflects the uniformity of temperature distribution within the electric furnace; a larger value indicates a more uneven temperature distribution. The power timing allocation unit, based on smelting process requirements and predicted thermal field trends, divides the entire smelting process into three stages: preheating, smelting, and holding, and calculates the appropriate power allocation ratio for each stage. The efficiency optimization unit analyzes the current temperature field characteristics and energy consumption data to calculate an energy efficiency improvement index, which reflects the differences in energy utilization under different power allocation schemes.

[0073] The power parameter synthesis unit integrates the calculation results from the above three units and uses a weighted fusion method to generate the final power control parameter set. This fusion process can be expressed by the following formula:

[0074]

[0075] in: This represents the final set of power regulation parameters. Indicates the regional temperature difference index. Represents the power phase allocation index. This is an energy efficiency improvement indicator. Coefficient. , and These are weighted parameters that are dynamically adjusted based on actual operating conditions, reflecting the relative importance of temperature uniformity, stage matching degree, and energy efficiency optimization in power allocation. These weighted parameters are adaptively adjusted according to the characteristics of smelting materials, process requirements, and real-time operating status.

[0076] After receiving the power control parameter set, the cooling dynamic adaptation module first extracts the adsorption characteristics of the medium on the furnace surface. These characteristics include parameters such as the coverage area, adhesion strength, and distribution uniformity of the cooling medium on the furnace surface. The module then incorporates environmental variables, such as ambient temperature, humidity, and air pressure, to predict the diffusion characteristics of the cooling medium on the furnace surface. The diffusion characteristic analysis considers factors such as the physical properties of the medium, its flow state, and surface interactions, generating an adsorption-diffusion parameter set that includes parameters such as diffusion rate, penetration depth, and uniformity.

[0077] Based on the adsorption-diffusion parameter set, the cooling dynamic adaptation module calculates the optimal cooling path using fluid dynamics principles. The calculation process considers the interplay of multiple variables, such as medium flow velocity, injection angle, and coverage area, and finds the best parameter combination through iterative optimization. The final cooling path control result includes specific control parameters such as nozzle opening adjustment value, flow distribution ratio, and injection direction correction. These parameters are transmitted in real time to the cooling system actuators, achieving dynamic optimization of the cooling process.

[0078] A two-way feedback mechanism is established between the power allocation strategy module and the cooling dynamic adaptation module. Changes in power allocation immediately trigger the recalculation of cooling parameters, while changes in cooling effect, in turn, affect the power allocation strategy. This closed-loop control method enables the system to respond quickly to changes in operating conditions and maintain the stability of the electric furnace operation. The two modules share the same state evaluation model, which continuously monitors the combined effect of power allocation and cooling regulation, providing a reference for the next control decision.

[0079] In terms of implementation technology, the power allocation strategy module adopts a combination of rule-based reasoning and machine learning. For known typical operating conditions, a predefined rule base is used to quickly generate a power allocation scheme; for complex or novel operating conditions, a trained neural network model is invoked for calculation. This approach ensures both fast response in common situations and retains flexibility in handling special cases. The cooling dynamic adaptation module mainly relies on a combination of computational fluid dynamics simulations and experimental data, improving control accuracy by establishing an accurate medium flow model.

[0080] The module's data processing flow is meticulously designed to meet the requirements of real-time control. Temperature field data undergoes noise reduction and feature extraction before being fed into a fast decision engine. Decision results undergo a safety check before being sent to the actuators to ensure no dangerous control commands are generated. The latency of the entire processing chain is strictly controlled within a certain range to meet the timeliness requirements of real-time control.

[0081] During system operation, the power allocation strategy module continuously records the implementation effects of different control schemes, forming an experience database. This data is used to periodically update the decision-making model, enabling the system to continuously optimize control strategies. The cooling dynamic adaptation module also possesses learning capabilities, gradually improving the medium flow model and enhancing prediction accuracy by analyzing historical control records.

[0082] The collaborative operation of the two modules enables multi-dimensional optimized control of the medium-frequency induction furnace's operating status. The power distribution strategy module ensures a reasonable temperature field distribution from an energy input perspective, while the cooling dynamic adaptation module supplements and adjusts the temperature from a heat dissipation perspective. This dual control mechanism effectively improves the overall efficiency of the furnace operation while reducing the probability of abnormal operating conditions. The system's adaptive capability enables it to cope with uncertainties such as changes in raw materials and environmental fluctuations, maintaining stable control performance.

[0083] In actual deployment, the module adopts a layered computing architecture. The core algorithm runs on high-performance computing nodes, while the real-time control part is deployed on edge computing units closer to the device. This architecture satisfies the computational requirements of complex algorithms while ensuring the timely issuance of control commands. The system also features a robust exception handling mechanism that automatically switches to a safe operating mode when sensor malfunctions or communication interruptions are detected, preventing equipment damage due to control system problems.

[0084] The performance of the power distribution strategy module and the cooling dynamic adaptation module relies on accurate temperature field data and reliable predictive information. Therefore, the modules incorporate data quality detection capabilities to identify and filter abnormal data, preventing erroneous data from influencing control decisions. Simultaneously, the modules establish close collaboration with other parts of the platform to jointly maintain the stability and safety of system operation. Through this systematic design, the two modules achieve intelligent, precise, and reliable control of the medium-frequency induction furnace operation.

[0085] Example 4: The abnormal state intervention module and the operational performance evaluation module constitute the intelligent control core of the medium-frequency electric furnace operation monitoring platform, ensuring stable furnace operation through real-time monitoring and dynamic adjustment. Based on temperature fluctuation characteristics and thermal field offset analysis, the system achieves accurate identification and proactive intervention of abnormal operating conditions, while continuously optimizing operating parameters to improve overall efficiency.

[0086] Under typical operating conditions, the system detected an abnormal change in the temperature fluctuation rate in area B. Data collected by the temperature sensor array showed that the temperature gradient in this area changed significantly higher than the historical average over three consecutive sampling periods. The module immediately initiated a multi-dimensional analysis program, spatially comparing the real-time temperature distribution with the predicted trajectory to calculate the thermal field offset. Data analysis indicated that the anomaly stemmed from a decrease in local heat dissipation efficiency caused by uneven distribution of the cooling medium. The system then retrieved historical case databases, matched processing records under similar operating conditions, and generated an intervention plan that included adjustments to the medium flow rate and optimization of the injection angle.

[0087] The correspondence between temperature fluctuation characteristics and intervention parameters is established based on the analysis of a large amount of operational data. Table 1 shows the intervention strategies adopted by the system according to different types of fluctuation characteristics. These strategies have been verified to have good adaptability through long-term operation.

[0088] Table 1: Correspondence between Temperature Fluctuation Characteristics and Intervention Strategies

[0089]

[0090] For the case in Area B mentioned above, the system identified it as exhibiting a continuously rising fluctuation characteristic. Based on the strategy framework in Table 1, the module executed a phased adjustment of the medium flow rate: initially increasing the flow rate to an appropriate multiple of the baseline value, and then fine-tuning based on real-time feedback. Simultaneously, the power compensation scheme was implemented in stages, first stabilizing the power output of adjacent areas, and then gradually adjusting the power supply parameters of the abnormal area. This step-by-step approach avoided drastic system oscillations and achieved a smooth transition.

[0091] The operational performance evaluation module was simultaneously activated for tracking and analysis. The system recorded changes in key parameters during the intervention process, including temperature stabilization time, energy consumption fluctuation curves, and regional temperature difference changes. By establishing a multi-dimensional performance evaluation model, the system found that appropriately widening the temperature control range in zone B improved overall stability, while significantly reducing energy consumption fluctuations in adjacent areas. These findings were integrated into the electric furnace operational performance data table as a reference for optimization and adjustment.

[0092] During anomaly handling, the system employs a distributed computing architecture to ensure real-time response. Edge computing nodes are responsible for basic data collection and rapid intervention command issuance, while the central processing unit performs in-depth analysis and strategy optimization. This architecture enables the system to complete the entire process from anomaly detection to intervention implementation within milliseconds. Simultaneously, all anomaly events and processing records are stored in a historical database, providing material for the system's self-learning.

[0093] The performance evaluation module employs sliding time window analysis technology to dynamically calculate operational indicators for different time periods. The system analyzes the correlation between temperature stability and energy consumption on a production cycle basis, identifying potential optimization opportunities. Evaluation results are presented visually to help operators intuitively understand the system's operating status and optimization directions.

[0094] In a typical application, the system successfully handled a multi-regional cascading anomaly. A sudden temperature surge occurred in zone C, triggering fluctuations in the adjacent zone D. Through real-time thermal field analysis, the system identified the anomaly propagation path and immediately activated a comprehensive anomaly handling scheme. The media flow rate was adjusted in a coordinated manner across regions, and a dynamic balancing strategy was implemented for power compensation, restoring the system to a stable state within a short time. Post-event analysis showed that this comprehensive intervention strategy effectively prevented the anomaly from spreading and avoided potential production interruptions.

[0095] The system's knowledge base continuously accumulates handling experience. After each intervention, the system automatically generates a case report, recording the anomaly characteristics, handling measures, and final results. These cases undergo cluster analysis and feature extraction to form more comprehensive response strategies. Simultaneously, the effectiveness evaluation results are fed back into the knowledge base to optimize future decision-making.

[0096] The human-computer interface provides operators with a complete monitoring view. When an anomaly occurs, the system highlights the affected area and uses color coding to distinguish the severity level. During the implementation of the intervention plan, real-time effect feedback is displayed in the form of trend curves, facilitating manual supervision and intervention adjustments when necessary. All operation records and system decisions generate detailed logs, supporting post-event review and analysis.

[0097] The system's adaptive capability is reflected in its dynamic parameter adjustment. As operating time accumulates, the modules gradually optimize the threshold settings for each region, making them more aligned with actual operating conditions. Performance evaluation criteria are also adjusted accordingly based on changes in production processes, ensuring the applicability of the evaluation results. This continuous evolutionary characteristic enables the system to maintain optimal operating conditions over the long term.

[0098] In real-world operating environments, the module needs to handle various complex situations. For example, when a sensor experiences a temporary malfunction, the system can compensate for the data using data from adjacent measuring points and historical patterns to maintain basic monitoring functions. For new abnormal modes, the system will attempt conservative interventions while ensuring safety, and record the processing results for optimization.

[0099] Example 5: The dynamic verification module is an important component of the medium-frequency electric furnace operation monitoring platform. It is responsible for real-time verification and anomaly handling of the multi-dimensional temperature distribution codes generated by the temperature field modeling module. This module ensures the accuracy and reliability of temperature field data by establishing a multi-level verification mechanism, providing a high-quality data foundation for analysis and decision-making.

[0100] The core function of this module is to monitor the multi-dimensional temperature distribution encoding output by the temperature field modeling module in real time. The system continuously scans the feature values ​​of each region in the encoding and compares them with the preset normal operating threshold range. When the feature value of a certain region exceeds the threshold, the module immediately initiates the anomaly handling process. First, it extracts the map data of the corresponding sub-region from the original thermal image. This data retains the original temperature distribution information of the region and is unaffected by the feature extraction and encoding process. The sub-region map contains detailed information such as the temperature gradient distribution, hotspot locations, and boundary changes of the region.

[0101] The extracted sub-region maps are sent to the anomaly intervention module for secondary analysis. This secondary analysis employs more sophisticated algorithms, focusing on the temperature conduction relationship, heat diffusion characteristics, and trends over time between the anomaly region and its surrounding areas. The analysis incorporates historical performance data of the region to assess the severity and development trend of the current anomaly. Based on the results of the secondary analysis, the dynamic verification module generates targeted parameter update suggestions, which are used to correct and improve the original anomaly intervention parameter set.

[0102] The parameter update process employs a gradual adjustment strategy. The system does not immediately replace all existing parameters with new ones; instead, it determines an appropriate parameter mixing ratio based on the nature and severity of the anomaly. For sudden, severe anomalies, the new parameters have a larger weight; for gradual, minor anomalies, a smaller proportion of parameters is updated. This strategy ensures both a rapid response to severe anomalies and avoids overreacting to minor fluctuations.

[0103] The dynamic verification module has established a close collaborative relationship with other modules on the platform. After parameter updates are completed, the module feeds back the adjustment information to the temperature field modeling module, prompting it to adaptively adjust the feature extraction algorithm. Simultaneously, the updated anomaly intervention parameters are shared with the trajectory prediction module, helping it improve the accuracy of the thermal field evolution model. This comprehensive collaborative mechanism enables the entire system to continuously optimize the collaborative working effect of its various modules.

[0104] In terms of implementation technology, the dynamic verification module adopts a distributed computing architecture. The threshold monitoring function is deployed on edge nodes close to the data source to achieve millisecond-level anomaly detection; the secondary analysis function runs on a central server with stronger computing power to complete complex pattern recognition and decision generation. This architecture design satisfies both real-time requirements and ensures the quality of analysis and decision-making.

[0105] The module's anomaly detection algorithm is built upon machine learning technology. By analyzing a large amount of historical operating data, the system automatically learns the normal fluctuation range of temperature characteristic values ​​in different areas under various operating conditions. The detection model is updated regularly to reflect the impact of changes in equipment status and process adjustments. For new anomaly patterns, the module has online learning capabilities, allowing it to quickly expand its recognition range.

[0106] The secondary analysis employed a multi-scale analysis method. Spatially, it examined both the local characteristics of the anomalous region and its positional relationship within the overall temperature field; temporally, it analyzed the development rate and duration of the anomaly. This multi-dimensional analysis method enables a more accurate assessment of the anomaly's nature and potential impact.

[0107] The parameter update mechanism incorporates a robust security verification process. All generated update suggestions undergo consistency checks to ensure they align with physical laws and process requirements. Significant parameter adjustments require manual verification before implementation. This cautious update strategy effectively prevents system instability caused by erroneous adjustments.

[0108] The dynamic verification module boasts excellent scalability. New detection algorithms and analysis methods can be easily integrated into the system without affecting the normal operation of existing functions. The module's interface design supports interfacing with different types of temperature sensing devices and control systems, adapting to diverse industrial application scenarios.

[0109] In actual operation, the module needs to handle various complex situations. For example, when anomalies occur simultaneously in multiple areas, the module will assess the correlation between the anomalies and determine the priority and coordination plan for handling them. Another example is during process changeovers, where the module can automatically adjust detection thresholds to adapt to new operating conditions.

[0110] The system's fault tolerance has been carefully designed. When some sensor data is missing or communication is interrupted, the module can make reasonable inferences based on existing information and maintain basic verification functions. For updates to key parameters, the system retains multiple historical versions and can quickly revert to a previous stable state if necessary.

[0111] Performance optimization of the dynamic verification module is an ongoing process. The system records detailed data for each anomaly handling, including detection time, analysis results, and parameter update effects. This data is used to periodically evaluate the module's performance, guiding algorithm improvements and parameter adjustments.

[0112] The module's human-computer interface design takes into account the actual needs of operators. Anomaly information is presented intuitively, including area location markers, anomaly severity assessments, and handling suggestions. Operators can easily view detailed information and adjust automatic processing parameters as needed.

[0113] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0114] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A medium-frequency electric furnace operation monitoring platform, characterized in that, include: The temperature field modeling module is used to generate multi-dimensional temperature distribution codes based on the real-time thermal imaging of the medium-frequency electric furnace. The trajectory prediction module predicts the thermal field evolution trajectory during the operation of the electric furnace based on the initial temperature code in the multi-dimensional temperature distribution code and the initial thermal imaging map. The power allocation strategy module generates a power control parameter set based on the current temperature code in the multi-dimensional temperature distribution code and the predicted thermal field evolution trajectory. The cooling dynamic adaptation module analyzes the flow characteristics of the cooling medium on the furnace surface based on the power regulation parameter set and generates cooling path regulation results. The abnormal state intervention module extracts temperature fluctuation characteristics and thermal field offset based on the cooling path control results, and generates an abnormal intervention parameter set. The operation performance evaluation module analyzes the target temperature zone distribution and stability duration based on the abnormal intervention parameter set, and generates an electric furnace operation performance data table. The trajectory prediction module superimposes a disturbance factor on the initial temperature code, and learns to remove the disturbance using the initial thermal imaging image and corresponding smelting process parameters as conditions, and predicts the thermal field evolution trajectory during the operation of the electric furnace. The power allocation strategy module includes: a thermal field equilibrium assessment unit, which calculates the regional temperature difference index based on the current temperature code and the thermal field evolution trajectory; The power timing allocation unit calculates the power stage allocation index based on the current temperature code and the thermal field evolution trajectory. The efficiency optimization unit generates energy efficiency improvement indicators based on the current temperature encoding. The power parameter synthesis unit integrates the regional temperature difference index, power stage allocation index, and energy efficiency improvement index to generate a power regulation parameter set; The cooling dynamic adaptation module executes the following steps: extracting the adsorption characteristics of the medium on the furnace surface based on the power control parameter set; predicting the diffusion characteristics of the medium by combining environmental variables to generate an adsorption diffusion parameter set; and adjusting the medium flow rate distribution ratio according to the adsorption diffusion parameter set to generate the cooling path control result. The abnormal state intervention module execution process is as follows: Based on the cooling path control results, monitor the temperature fluctuation rate and thermal field offset. Analyze the impact of fluctuation range on metallurgical reaction rate to generate reaction rate intervention data; Based on the reaction rate intervention data, the medium distribution path in the target temperature zone is dynamically adjusted to generate an abnormal intervention parameter set.

2. The medium-frequency electric furnace operation monitoring platform according to claim 1, characterized in that, The temperature field modeling module consists of a feature extraction unit and a field reconstruction unit. The feature extraction unit decomposes the thermal image into discrete temperature feature vectors, and the field reconstruction unit converts the discrete temperature feature vectors into a temperature field distribution map to generate a multi-dimensional temperature distribution code.

3. The medium-frequency electric furnace operation monitoring platform according to claim 2, characterized in that, The temperature field modeling module performs a training process before generating a multi-dimensional temperature distribution code: historical thermal images are collected and input into the feature extraction unit to generate temperature feature vectors; The temperature feature vector is reconstructed into a temperature field distribution map through the field reconstruction unit; The differences between the reconstructed map and the actual thermal image are compared and the network parameters are updated until convergence.

4. The medium-frequency electric furnace operation monitoring platform according to claim 1, characterized in that, Before predicting the thermal field evolution trajectory, the trajectory prediction module optimizes the network parameters by maximizing the similarity of the probability distribution of the thermal field evolution trajectory.

5. The medium-frequency electric furnace operation monitoring platform according to claim 1, characterized in that, The operational performance evaluation module execution process is as follows: Based on the abnormal intervention parameter set, target temperature zone distribution data are collected at different time periods; Analyze the temperature range stability duration and energy consumption trends to generate target temperature range performance parameters; Based on the performance parameters of the target temperature zone, adjust the temperature control range to generate an electric furnace operation performance data table.

6. The medium-frequency electric furnace operation monitoring platform according to claim 1, characterized in that, It also includes a dynamic verification module: when the feature value of a specific region in the multi-dimensional temperature distribution code generated by the temperature field modeling module exceeds the threshold, the corresponding sub-region map is extracted from the thermal imaging map; The sub-region map is input into the abnormal state intervention module for secondary analysis; The abnormal intervention parameter set is updated based on the results of the secondary analysis.

Images