AI Construction Control Method and System for Unshaped Refractory Materials for Submerged Arc Furnaces

By acquiring environmental and material data of the electric arc furnace construction area and adjusting construction parameters using a construction quality prediction model, the problems of construction status differences and model input deviations at the overlapping boundaries during construction were solved, thus achieving data continuity and control stability in the construction process.

CN123113553APending Publication Date: 2026-11-17BEIJING MINGYIN TECHNOLOGY CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611408633.5
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-09-11
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

During the construction of the lining of the electric arc furnace, the construction quality prediction model is difficult to continuously participate in the construction decision-making process. The construction status at the overlap boundary is different from that inside the area, and the model lacks a safety control boundary when the input deviates from the feature space of the historical training samples.

Method used

By acquiring environmental conditions and material performance data of the current construction area, the construction mix ratio and process parameters are determined using a construction quality prediction model. The parameters are then adjusted based on the actual construction response, and the construction parameters at the overlap boundary are corrected. Safety protection parameters are adopted to ensure construction quality.

Benefits of technology

It achieves data continuity and control stability in the construction process, improves the consistency between the overlapping position and the internal construction status of the area, and ensures that the applicability of the construction quality prediction model matches the actual construction status.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123113553A_ABST
    Figure CN123113553A_ABST
Patent Text Reader

Abstract

This invention relates to the field of industrial process automation control and refractory material construction technology, and discloses an AI-based construction control method and system for unshaped refractory materials used in submerged arc furnaces. The construction control unit acquires environmental and material performance data, uses a construction quality prediction model to evaluate candidate mix proportions, and determines the construction mix proportions and parameters for the placement, pouring, and compaction processes. Corresponding process monitoring data forms a construction status evaluation result, which is used to determine parameters for the next construction area. When the model input deviates from the feature space of historical training samples, safety degradation control is implemented. This invention enhances the data continuity and safety of cross-process and cross-regional construction control in the continuous construction of submerged arc furnace linings.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of industrial process automatic control and refractory material construction technology, and discloses an AI construction control method and system for unshaped refractory materials used in submerged arc furnaces. Background Technology

[0002] The lining of an electric arc furnace is subjected to high temperatures, thermal circulation, material erosion, and chemical corrosion for extended periods. The structural condition of the lining formed during construction directly affects its subsequent service stability. Unshaped refractory materials are formed on-site according to the construction location, and the construction result is simultaneously affected by the material composition, moisture content, flow properties, construction environment, and the execution of the placement, pouring, and compaction processes.

[0003] Current construction control methods primarily involve monitoring and adjusting feed rate, construction thickness, pouring flow rate, pouring pressure, or compaction parameters separately. When mix design, material placement, pouring, and compaction are independent of each other, the material condition and the actual construction results from the previous process are difficult to continuously influence subsequent construction decisions, and data transmission between processes is prone to interruption.

[0004] Building a construction quality prediction model using historical construction data can predict the construction quality status based on construction conditions. However, when candidate mix proportions are not used as input to the model, it is difficult for the model to establish a clear correspondence between different candidate mix proportions and the construction quality prediction results; when there is a lack of feedback between the mix proportion determination results and the actual construction response on site, the model prediction results are also difficult to participate in the subsequent parameter adjustment of the area as the construction status changes.

[0005] The lining of an electric arc furnace is typically constructed continuously in multiple zones. After the previous zone is completed, the edge material has already undergone a certain period of hardening and setting before the next zone begins construction. The flow and deformation capabilities of the material at the overlap boundary differ from those of the newly applied material. If the same material placement and compaction parameters are used at the overlap location, the construction condition at the overlap boundary is likely to differ from that within the zone.

[0006] The flow rate and pressure response during the pouring process can reflect the actual material transport resistance, and the actual thickness obtained during the spreading process can reflect the current laying status of the area. When flow rate, pressure, and spreading thickness are not further involved in determining the compaction process parameters, there is a lack of direct data link between the compaction parameters and the material's rheological state in the field.

[0007] The reliability of construction quality prediction models is also limited by the coverage of the feature space of historical training samples. When the current construction status deviates significantly from the feature space of historical training samples, continuing to directly adjust the construction mix ratio and construction parameters based on the model output will not allow the model's control authority to match its applicable scope. Summary of the Invention

[0008] The purpose of this invention is to provide an AI construction control method and system for unshaped refractory materials used in submerged arc furnaces. This system addresses the problems of lacking continuous data relationships between mix proportion determination, process execution, and subsequent construction area parameter adjustments when material properties, construction environment, overlapping status of adjacent areas, and actual process responses change during the construction of submerged arc furnace linings, as well as the lack of corresponding safety control boundaries when the model input deviates from the feature space of historical training samples.

[0009] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows:

[0010] AI construction control methods for monolithic refractories used in submerged arc furnaces include:

[0011] Obtain environmental status data of the current construction area and perform performance testing on the unshaped refractory material to be constructed to obtain material performance status data;

[0012] The environmental status data, material performance status data, construction requirements of the current construction area, and candidate mix characteristics corresponding to different candidate mix ratios are input into the construction quality prediction model to obtain the construction quality prediction results corresponding to each candidate mix ratio. Based on the construction quality prediction results corresponding to each candidate mix ratio, the construction mix ratio of the current construction area is determined, and the initial construction parameters for the material placement, pouring, and compaction processes are formed.

[0013] The material placement, pouring, and compaction are carried out according to the initial construction parameters. The target compaction energy density of the compaction process is determined based on the material flow response of the pouring process and the actual material thickness of the material placement process. The construction parameters are adjusted according to the monitoring data of each process.

[0014] By linking the monitoring data of the processes of material placement, pouring, and compaction, a construction status evaluation result for the current construction area is formed;

[0015] The construction status evaluation results, along with the environmental status data and material performance data of the next construction area, are used to determine the construction parameters of the next construction area. The material placement and compaction parameters at the overlap position are then adjusted based on the overlap boundary temperature and the degree of hardening of the edge material.

[0016] When the model input deviates from the feature space of historical training samples, the construction quality prediction model is restricted from modifying the construction mix ratio, and safety protection construction parameters are adopted; the process is executed cyclically according to the construction area sequence until the predetermined construction area is completed.

[0017] As a preferred technical solution of the present invention, the environmental status data includes at least multiple data such as ambient temperature, ambient humidity, airflow status and furnace shell temperature in the construction area; the data collected from different locations are filtered, calibrated and time-aligned, and an environmental status dataset is formed according to the corresponding construction area.

[0018] As a preferred technical solution of the present invention, the material performance status data includes at least the chemical composition, water content and flow properties of the unshaped refractory material; representative samples are obtained from the material to be constructed, and the material performance status data is obtained through composition detection, water content detection and flow property detection, and the detection results are associated with the corresponding material batch and construction area.

[0019] As a preferred technical solution of the present invention, the construction quality prediction model is established based on historical construction records, which include at least environmental conditions, material performance conditions, construction mix proportions, construction parameters, and post-construction quality inspection results. During training, the data corresponding to environmental conditions, material performance conditions, construction requirements, and construction mix proportions are used as model inputs, the corresponding quality inspection results are used as supervisory information, and the training objective is to reduce the prediction error between the construction quality prediction results and the corresponding quality inspection results, so that the construction quality prediction model establishes a mapping relationship between construction conditions, candidate mix proportions, and construction quality.

[0020] As a preferred technical solution of the present invention, the candidate mix proportions that meet the material type and construction requirements are obtained from the mix proportion database, and the mix proportion composition information corresponding to each candidate mix proportion is input as the candidate mix proportion feature into the construction quality prediction model.

[0021] The construction control unit forms the current input feature vector using the same data processing method as the training phase of the construction quality prediction model. Based on the covariance relationship between the input features of the historical training samples, it normalizes the multivariate deviation between the current input feature and the feature center of the historical training samples to obtain the distance evaluation result of the current input relative to the feature space of the historical training samples.

[0022] The confidence index of the construction quality prediction result is determined based on the correspondence between the distance evaluation results and the construction quality prediction error described in the historical training phase. When the confidence index is lower than the preset safety threshold, the construction quality prediction model stops its autonomous modification of the construction mix ratio for the next construction area, maintains the construction mix ratio that is currently being executed normally, and calls the expert rule library of refractory material rheological limits to generate safety protection construction parameters.

[0023] As a preferred technical solution of the present invention, when performing material placement, the target material placement state is determined according to the target construction thickness of the current construction area. The actual material placement thickness and actual material placement amount are obtained by the thickness detection device and material placement amount detection device set in the automatic material placement device. The feeding speed and the working state of the leveling mechanism are adjusted according to the deviation of the actual material placement thickness and actual material placement amount from the target material placement state. The adjusted material placement parameters and corresponding monitoring results are sent to the construction status evaluation process.

[0024] As a preferred technical solution of the present invention, during the pouring process, the actual pouring volume flow rate is obtained through the flow detection unit, the pipeline delivery pressure difference is obtained through the pressure detection units set at both ends of the pressure detection section, the driving state of the material conveying mechanism is adjusted according to the changes of the actual pouring volume flow rate and the pipeline delivery pressure difference relative to the corresponding target state, and the actual pouring volume flow rate and the pipeline delivery pressure difference are sent to the construction status evaluation process and the compaction parameter determination process.

[0025] As a preferred embodiment of the present invention, during the compaction process, the equivalent apparent viscosity used to characterize the actual transport resistance of the current material is determined based on the actual pouring volume flow rate monitored during the pouring process and the pipeline transport pressure difference at both ends of the pressure detection section, combined with the pipeline structure parameters of the pressure detection section. Furthermore, the target compaction energy density of the current construction area is determined according to the following relationship, based on the actual material thickness obtained during the material placement process:

[0026] ;

[0027] in, The target compaction energy density is expressed in J / m³. This is the energy density correction factor corresponding to the material type, in units of s⁻¹; The equivalent apparent viscosity is determined based on the actual pouring volume flow rate, pipeline delivery pressure difference, and pipeline structural parameters of the pressure testing section, and is expressed in Pa·s. This refers to the actual fabric thickness, in meters (m). The target construction thickness is expressed in meters (m). The thickness correction index corresponds to the material type and is dimensionless.

[0028] The compaction execution unit is based on the target compaction energy density. The calibration relationship of the compaction equipment is adjusted to control the compaction force, vibration frequency, and single-point compaction duration. The displacement response and stress response of the material are obtained through displacement detection device and pressure detection device.

[0029] As a preferred technical solution of the present invention, when forming the construction status evaluation result, the thickness distribution state formed by the material placement process, the flow and pressure response state formed by the pouring process, and the displacement and stress response state formed by the compaction process are associated according to the corresponding construction areas, and the construction status evaluation result is determined based on the deviation of the actual response of each process from the target construction state of the current construction area.

[0030] As a preferred technical solution of the present invention, when the construction status evaluation results are used for the selection of mix proportions and determination of process parameters in the next construction area, the overlap boundary temperature between the current construction area and the next construction area and the initial setting and hardening degree of the edge material of the current construction area are obtained.

[0031] Based on the current material's setting and hardening characteristics, the initial setting and hardening degree and the overlap boundary temperature normalized by the reference temperature are combined to convert into an interfacial rheological attenuation coefficient. This interfacial rheological attenuation coefficient represents the decrease in the residual flow and deformation capacity of the edge material as the setting and hardening degree of the edge material increases. The initial material thickness and compaction parameters near the overlap boundary in the next construction area are then corrected based on the interfacial rheological attenuation coefficient.

[0032] As a preferred technical solution of the present invention, during the construction process, environmental status data, material performance status data, and operating data of each construction equipment are continuously monitored. When the detection results exceed the allowable range of the corresponding equipment or construction process, at least one abnormal control operation is performed according to the degree of abnormality, such as stopping the corresponding construction process, reducing the construction speed, maintaining the current safety status, and outputting alarm information. The construction area and construction status data corresponding to the occurrence of the abnormality are recorded.

[0033] As a preferred technical solution of the present invention, the environmental status data, material performance status data, actual construction mix ratio, actual construction parameters of each process, process monitoring data and construction quality test results after the completion of construction are associated with the construction area and written into the construction database. The construction quality prediction model is updated using the newly added construction records with construction quality test results. After the updated model is verified based on historical construction records, it is used for construction quality prediction in subsequent construction areas.

[0034] The present invention also discloses an AI construction control system for monolithic refractory materials for submerged arc furnaces, which is used to execute the AI ​​construction control method for monolithic refractory materials for submerged arc furnaces as described above, including an environmental state perception unit, a material performance detection unit, a construction control unit, a material placement execution unit, a pouring execution unit, and a compaction execution unit;

[0035] The environmental condition sensing unit includes an environmental monitoring device arranged in the construction area of ​​the lining of the electric arc furnace, which is used to collect environmental condition data and send it to the construction control unit.

[0036] The material performance testing unit includes a material sampling device and a testing device corresponding to the material performance parameters, which is used to obtain the material performance status data of the unshaped refractory material to be constructed and send it to the construction control unit.

[0037] The construction control unit includes an edge computing terminal and a cloud server. The edge computing terminal is deployed at the construction site of the electric arc furnace and is connected to the environmental condition sensing unit, material performance testing unit, concrete placement execution unit, pouring execution unit, and compaction execution unit. It is used to call the local construction quality prediction model, generate construction control parameters, and receive process monitoring data. The cloud server is used to store the construction database and to offline train and update the construction quality prediction model based on new construction records.

[0038] The material placement execution unit, pouring execution unit, and compaction execution unit each include a controller, a construction execution mechanism, and sensors for detecting the actual response of the corresponding process. The controller drives the construction execution mechanism according to the construction parameters issued by the edge computing terminal and feeds back the process monitoring data to the edge computing terminal.

[0039] The edge computing terminal is connected to the cloud server through an industrial IoT gateway with data interruption and resume function; when the communication network is interrupted, the edge computing terminal independently completes the feedback control between construction areas based on the locally cached construction quality prediction model parameters and construction control data.

[0040] Compared with the prior art, the present invention has the following beneficial effects:

[0041] This invention employs a control method that integrates environmental state data, material performance data, construction requirements, and candidate mix proportion characteristics to predict construction quality. This allows different candidate mix proportions to correspond to specific construction quality prediction results, and the determined mix proportions are then used to determine parameters for the placement, pouring, and compaction processes, preventing a disconnect between mix proportion evaluation and on-site construction execution. The actual placement thickness and quantity of material formed during the placement process, the actual pouring volume flow rate and pipeline pressure difference formed during the pouring process, and the displacement and stress responses formed during the compaction process are correlated according to the construction area. The construction state evaluation results continue to participate in the mix proportion selection and construction parameter determination for the next construction area, ensuring that the actual construction response of the previous construction area influences subsequent construction decisions. The material transport response obtained during the pouring process and the actual placement thickness obtained during the placement process are used together to determine the target compaction energy density, allowing compaction parameters to vary with actual material transport resistance and laying conditions. The overlap boundary temperature and the degree of hardening of the edge material are used to form the interfacial rheological attenuation coefficient, which is then used to correct the material placement and compaction parameters near the overlap boundary in the next construction area, thus improving the inconsistency between the overlap location and the internal construction state of the area during continuous zone construction. During the operation of the construction quality prediction model, the confidence level of the prediction results is determined by evaluating the deviation of the current input from the feature space of historical training samples. If the confidence level is insufficient, the model is restricted from autonomously modifying the construction mix design and safety protection parameters are adopted to ensure that the model's control authority matches its applicable state. Edge computing terminals handle on-site construction control, while cloud servers are used for construction database storage and updates to the construction quality prediction model. Even when the communication network is interrupted, construction control can continue based on local caching, thus maintaining data continuity and control stability during the construction process. Attached Figure Description

[0042] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained from these drawings without creative effort.

[0043] Figure 1 A flowchart illustrating the overall process of AI-based construction control for monolithic refractory materials used in submerged arc furnaces.

[0044] Figure 2 A flowchart illustrating the process of generating environmental condition data and material performance condition data.

[0045] Figure 3 A flowchart for evaluating candidate mix proportions and determining construction mix proportions.

[0046] Figure 4This is a flowchart of the fabric processing control process.

[0047] Figure 5 This is a flowchart for controlling the pouring process.

[0048] Figure 6 This is a flowchart for controlling the compaction process.

[0049] Figure 7 This is a flowchart of the construction status evaluation and parameter adjustment process for the next construction area.

[0050] Figure 8 A flowchart for the interface rheological compensation process.

[0051] Figure 9 This is a flowchart of the model confidence assessment and security degradation control process. Detailed Implementation

[0052] In the following description, only certain exemplary embodiments are briefly described. As those skilled in the art will recognize, the described embodiments can be modified in various ways without departing from the spirit or scope of the embodiments of the invention. Therefore, the drawings and description are considered to be exemplary in nature and not restrictive.

[0053] The following is in conjunction with the appendix Figures 1-9 The embodiments of the present invention will be described in detail below.

[0054] Example 1: This example provides an AI-based construction control method for monolithic refractory materials used in submerged arc furnaces. The construction area for the submerged arc furnace lining is divided according to the furnace structure, the effective operating range of the construction equipment, and the construction organization requirements. Each construction area sequentially performs environmental condition acquisition, material performance testing, candidate mix design evaluation, construction parameter determination, material placement, pouring, compaction, and construction status evaluation. The construction status evaluation results generated in the current construction area are then used for the construction control of the next construction area.

[0055] Step S101: Generate environmental status data for the construction area. Temperature, humidity, airflow, and furnace shell temperature detection devices are installed in the construction area to obtain ambient temperature, ambient humidity, airflow status, and furnace shell thermal status, respectively. Each data point is recorded with its corresponding acquisition time and construction area identifier.

[0056] The construction control unit determines the validity of the detection data, filters and calibrates valid data, and aligns it according to a unified control cycle. The processed data from different sources collectively correspond to the current construction area, forming an environmental status dataset, which is used for candidate mix design evaluation and construction parameter determination.

[0057] Step S102: Generate material performance status data. Take samples from the current batch of unshaped refractory material to be constructed, test the material composition, moisture content and flow properties, and establish a correlation between the test results and the corresponding material batch and construction area.

[0058] When the same batch of materials is used continuously and the material condition does not change abnormally, the most recent valid material performance status data should be used for subsequent construction areas. If the material batch changes, the storage condition changes, or the actual construction response indicates that the material condition may have changed, resampling and testing should be conducted.

[0059] Material composition information, moisture content, and flow performance data are combined to form material performance status data, which is then sent to the construction control unit to participate in the evaluation of candidate mix proportions.

[0060] Step S103: Establish a construction quality prediction model. The construction quality prediction model is established based on historical construction records. A historical construction record used for model training must include at least environmental condition data, material performance data, actual construction mix proportions, concrete placement parameters, pouring parameters, compaction parameters, and post-construction quality inspection results.

[0061] The model training takes the characteristics corresponding to environmental conditions, material performance, construction requirements, and construction mix proportions as inputs, construction quality inspection results as supervisory information, and reduces the prediction error between the construction quality prediction results and the corresponding construction quality inspection results as the training objective, so that the construction quality prediction model can establish a mapping relationship between on-site construction conditions, construction mix proportions, and construction quality.

[0062] Model input features of different dimensions are processed according to the normalization or standardization parameters determined during the training phase. The same data processing parameters and feature order are used during model execution. After model training is complete, the model parameters and data processing parameters are stored in computing devices accessible to the construction control unit.

[0063] Step S104: Forming candidate mix proportions and performing mix proportion evaluation. The construction control unit retrieves candidate mix proportions that match the current material type and construction requirements from the mix proportion database, based on the current type of unshaped refractory material. The candidate mix proportions in the mix proportion database are derived from existing process mix proportion records.

[0064] For each candidate mix proportion, the construction control unit inputs the environmental status data of the current construction area, material performance status data, construction requirements, and mix proportion information of the corresponding candidate mix proportion into a model, and obtains the construction quality prediction results corresponding to different candidate mix proportions through the construction quality prediction model.

[0065] The construction control unit combines mix design constraints, construction equipment operating conditions, and construction quality prediction results to screen candidate mix designs and determine the mix design to be used in the current construction area.

[0066] When construction quality, construction time, and material usage status need to be considered simultaneously, multi-objective optimization is performed within a candidate range that meets the conditions for feasible mix proportions, equipment capacity, and construction safety. In some implementations, a non-dominated sorting genetic algorithm is used to search for adjustable mix proportion parameters and corresponding construction process parameters, and then construction schemes are selected based on construction quality prediction results and equipment operation constraints.

[0067] When the number of candidate mix proportions in the mix proportion database can be evaluated directly item by item, the construction quality prediction results corresponding to each candidate mix proportion can also be calculated and constrained for selection.

[0068] The construction control unit forms the current input feature vector according to the data processing method of the construction quality prediction model training stage, and uses the statistical characteristics of historical training samples to evaluate the degree of deviation of the current input state from the feature space of historical training samples.

[0069] During the historical training phase, the statistical centers of each input feature and the covariance relationships between features are preserved. After the current input is processed using the same standardization method as during the training phase, the construction control unit calculates the multivariate deviation of the current input from the statistical centers of the historical training samples, and then uses the covariance relationships between historical features to normalize and correct the deviations in different feature directions.

[0070] The resulting feature space distance reflects both the deviation of a single input feature from historical training samples and the difference between the combined state of multiple input features and historical training samples.

[0071] During the model validation phase, a correspondence is established between feature space distance and prediction error. During on-site operation, the construction control unit converts the current feature space distance into the confidence level of the construction quality prediction result according to the aforementioned correspondence; as the feature space distance increases, the confidence level decreases, and when the current input is located within the main coverage area of ​​historical training samples, the confidence level remains high.

[0072] The construction quality prediction results serve as input for evaluating candidate solutions, and the multi-objective optimization outputs construction mix proportions and construction process parameters. Data generated during on-site execution is written into the construction database and used for subsequent updates to the construction quality prediction model, ensuring a continuous data transfer relationship between model predictions, construction decisions, and actual execution results.

[0073] Step S105: Generate initial construction parameters for the process. The construction control unit obtains the initial construction parameters for material placement, pouring, and compaction from the corresponding construction process record based on the current material performance status, construction area requirements, and the construction mix ratio determined in step S104.

[0074] The initial construction parameters for fabric include the target fabric thickness and the corresponding feeding and leveling parameters; the initial construction parameters for pouring include the target material conveying state and the corresponding pouring drive parameters; and the initial construction parameters for compaction include the compaction parameters corresponding to the material type and construction thickness.

[0075] Initial construction parameters are sent to the material placement unit, pouring unit, and compaction unit, respectively.

[0076] Step S106: Perform the fabric application process. The fabric application unit includes a material storage structure, a feeding mechanism, a leveling mechanism, a thickness detection device, a fabric quantity detection device, and a local controller.

[0077] After receiving the target material placement status, the local controller drives the feeding actuator to place materials into the current construction area. The thickness detection device obtains the actual material thickness, and the material quantity detection device obtains the actual material output. The local controller adjusts the feeding and leveling states based on the deviation between the actual material thickness and the target material thickness.

[0078] When there is a significant discrepancy between the actual fabric thickness and the material output, the construction control unit records the abnormal status of the corresponding construction area. After fabric placement is completed, the actual thickness status, material output status, and actual fabric placement parameters are compiled into a fabric placement process record and sent to the construction control unit.

[0079] Step S107: Perform the pouring process. The pouring execution unit includes a material storage and mixing structure, a material conveying mechanism, a flow detection device, a pressure detection device, a pouring end, and a local controller.

[0080] The local controller drives the material conveying mechanism after receiving the target material conveying status. The flow detection device obtains the actual pouring volume flow rate, and the pressure detection units set at both ends of the pressure detection section obtain the inlet pressure and outlet pressure of the pressure detection section, respectively. The pipeline conveying pressure difference is obtained by subtracting the outlet pressure of the pressure detection section from the inlet pressure of the pressure detection section.

[0081] The construction control unit determines the material delivery status by combining the actual pouring volume flow rate and the pipeline delivery pressure difference. When the actual pouring volume flow rate is lower than the target status and the pipeline delivery pressure difference has not yet reached the equipment's allowable range boundary, the local controller adjusts the drive status of the material delivery mechanism.

[0082] When the pipeline pressure difference increases but the actual pouring volume flow rate does not increase synchronously, the corresponding state serves as the basis for judging abnormal transport resistance.

[0083] After the pouring is completed, the actual pouring volume flow rate, pipeline delivery pressure difference, and actual execution parameters are recorded to form a pouring process record and sent to the construction control unit.

[0084] Step S108: Perform the compaction process. The compaction execution unit includes a compaction actuator, a displacement detection device, a pressure detection device, and a local controller. The local controller drives the compaction actuator to act on the current construction area according to the compaction parameters issued by the construction control unit.

[0085] The construction control unit determines the equivalent apparent viscosity of the material based on the actual pouring volume flow rate, pipeline delivery pressure difference, and pipeline length and inner diameter of the pressure detection section obtained in step S107.

[0086] The equivalent apparent viscosity is determined according to the conversion relationship between flow rate, pressure difference, pipeline structural parameters, and material transport resistance established during the calibration of the pressure testing section equipment. For the same pressure testing section, the pipeline structural parameters remain fixed, and the equivalent apparent viscosity is updated during construction based on the actual pouring volume flow rate and pipeline transport pressure difference.

[0087] When the actual pouring volumetric flow rate decreases and the pipeline transport pressure difference increases, the determined equivalent apparent viscosity characterizes the increase in material transport resistance; when the actual pouring volumetric flow rate increases and the pipeline transport pressure difference decreases, the determined equivalent apparent viscosity characterizes the decrease in material transport resistance.

[0088] Equivalent apparent viscosity is used to convert the actual pouring volume flow rate and pipeline transport pressure difference under the same pressure testing conditions into a unified material response quantity that can participate in the compaction process control, and is not used as the true rheological constitutive parameter of unshaped refractories under all shear conditions.

[0089] When the actual pouring volume flow rate is 0, the equivalent apparent viscosity update is not performed. The current state is marked as having no effective flow rate, and the abnormal control logic determines whether to continue executing the corresponding construction area.

[0090] The construction control unit reads the actual fabric thickness formed during the fabric laying process. and the target construction thickness of the current construction area The target compaction energy density is determined according to the following relationship:

[0091] ;

[0092] in, The target compaction energy density for the current construction area is expressed in J / m³. This is the energy density correction factor corresponding to the current material type, in units of s⁻¹; The equivalent apparent viscosity is calculated based on the actual flow rate, pressure response, and calibration relationship of the pressure detection section during the pouring process, and is expressed in Pa·s. The actual thickness of the material laid in the current construction area, in meters (m). The target construction thickness for the current construction area, in meters (m). The thickness correction index is dimensionless and corresponds to the current material type.

[0093] and The parameters are determined based on the compaction calibration records for the corresponding material type and stored in the material performance parameter database. In one embodiment, Take 1.25s⁻¹, Take 1.5.

[0094] The target compaction energy density varies with the equivalent apparent viscosity, enabling the compaction process to respond to the actual material transport resistance. When the actual material thickness deviates from the target construction thickness, the thickness correction term further adjusts the target compaction energy density, allowing the compaction process to simultaneously consider the material flow state and the actual laying state.

[0095] The local controller determines the compaction force, vibration frequency, and single-point compaction duration within the allowable operating range of the equipment, based on the target compaction energy density and the relationship between the energy input and compaction parameters pre-calibrated by the corresponding compaction equipment.

[0096] The calibration relationship of the compaction equipment is obtained by calibrating the equipment under different compaction forces, vibration frequencies and durations of actual energy input, and stored in the local controller.

[0097] The displacement detection device acquires the displacement response during the compaction process, and the pressure detection device acquires the corresponding force response. The local controller combines the target compaction energy density, displacement response, and force response to determine the state change of the material under continuous compaction.

[0098] The compaction termination threshold is used to determine whether the displacement change of the current position has entered a stable state within multiple consecutive control cycles. The compaction termination threshold is the upper limit of the allowable displacement change, which is determined according to the compaction calibration record of the corresponding material and is associated with the material type and stored in the material performance parameter library.

[0099] When the displacement change of the current action position does not exceed the compaction termination threshold within multiple consecutive control cycles, and the actual cumulative input energy density determined according to the compaction equipment calibration relationship reaches the compaction state corresponding to the target compaction energy density, the compaction at the current action position ends or the device moves to the next action position.

[0100] If the displacement response has not yet reached the compaction requirements corresponding to the current material, and the equipment operating status is still within the allowable range, continue compaction or adjust the compaction parameters.

[0101] The actual cumulative input energy density is determined by the local controller based on the actual compaction parameters used during the compaction process, vibration frequency, single-point compaction duration, material volume corresponding to the current compaction area, and calibration data of the corresponding compaction equipment.

[0102] Thus, the actual flow rate and pressure response of the pouring process are transmitted to the compaction process via the equivalent apparent viscosity, and the actual material thickness formed in the material placement process is simultaneously involved in the determination of the target compaction energy density. A data transmission relationship based on the actual construction response of the material is formed among the three processes of material placement, pouring, and compaction.

[0103] Step S109: Generate the construction status evaluation result for the current construction area. The construction control unit associates the material placement process record, pouring process record, and compaction process record with the construction area identifier, and obtains the environmental status data, material performance status data, and actual construction mix ratio corresponding to the current construction area.

[0104] The actual thickness and material output status are extracted from the cloth laying process record; the actual pouring volume flow rate, pipeline conveying pressure difference, and the response relationship between the two are extracted from the casting process record; and the displacement response, stress response, and actual execution status are extracted from the compaction process record.

[0105] The construction control unit generates a construction status evaluation result for the current construction area based on the deviation of the actual response of each process from the corresponding target construction state. The allowable deviation range for each process is predetermined according to the corresponding material construction process requirements and the allowable operating range of the equipment.

[0106] After obtaining the quality inspection results after construction, the quality inspection results are then linked to the corresponding construction area records, so that the current construction area records can be used for the historical data accumulation of subsequent construction area control and construction quality prediction models.

[0107] Step S110: Implement feedback adjustments for the next construction area. Before entering the next construction area, re-obtain the environmental status data of the next construction area, and determine whether to re-perform material performance testing based on the material batch and material status.

[0108] The construction control unit reads the construction status evaluation results of the previous construction area. When the process response of the previous construction area meets the target state, the actual construction parameters can be used as a reference for the candidate construction parameters of the next construction area; when there is a process deviation in the previous construction area, the corresponding deviation state is used as feedback information for the construction decision of the next construction area, and the construction parameters that caused the deviation are not directly used.

[0109] For example, if the pipeline pressure difference increases and the actual pouring volume flow rate decreases in the previous construction area, the material delivery status should be re-evaluated in the next construction area based on the material flow properties, moisture content, and candidate mix proportions.

[0110] If the material thickness in the previous construction area meets the requirements but the compaction displacement response deviates from the compaction requirements, the corresponding construction parameters in the next construction area will be adjusted according to the material performance and compaction response.

[0111] Based on this, the construction control unit re-executes the candidate mix design evaluation and construction parameter determination, generating construction parameter adjustment instructions for the next construction area. The above process is repeated according to the predetermined construction sequence of the construction area until all predetermined construction areas have completed material placement, pouring, compaction, and construction status evaluation.

[0112] When adjacent construction areas are under continuous construction, the construction control unit performs rheological compensation of the overlap interface based on the overlap boundary temperature and the solidification and hardening state of the edge material of the current construction area.

[0113] The environmental condition sensing unit obtains the overlap boundary temperature between the current construction area and the next construction area through an infrared thermal imager. The time interval for completed compaction is determined based on the time when the current construction area is completed and the current time.

[0114] The construction control unit reads the setting and hardening characteristic data corresponding to the current unshaped refractory material. The setting and hardening characteristic data characterizes the relationship between the degree of setting and hardening of the corresponding material under different temperature and time conditions, and stores the corresponding setting and hardening degree data using temperature and time as indexes.

[0115] By utilizing the overlap boundary temperature and the time interval between completed compactions, neighboring data points are obtained from the current material's setting and hardening characteristic data. Interval interpolation is then performed between these neighboring data points to determine the initial setting and hardening degree of the material at the edge of the current construction area. .

[0116] Initial setting and hardening degree The value ranges from 0 to 1. An increase in the value indicates that the edge material has increased in degree of solidification and hardening relative to the moment of completion of compaction.

[0117] In one implementation, the following interfacial rheological attenuation relationship is established based on the current material solidification hardening characteristics:

[0118] ;

[0119] in, is the interfacial rheological attenuation coefficient, which is dimensionless and takes values ​​greater than 0 and less than or equal to 1. It is a natural exponential function; The rheological attenuation factor is a positive value corresponding to the current material's condensation characteristics and is dimensionless. This represents the initial hardening degree of the material at the edge of the current construction area, and is dimensionless. The absolute temperature of the overlap boundary, in Kelvin (K). The absolute temperature of the standard reference environment, in K;

[0120] rheological attenuation factor The parameters are determined based on the calibration results of the solidification hardening characteristics of the corresponding material and associated with the material type, and stored in the material performance parameter library.

[0121] The interfacial rheological attenuation coefficient converts the overlap boundary temperature and the degree of solidification and hardening of the edge material into an interfacial state quantity that can participate in the control of the next construction area, so that the time-related solidification state of the edge material of the previous construction area can continue to participate in the determination of construction parameters of the next construction area.

[0122] When the interface rheological attenuation coefficient decreases to the rheological compensation trigger range pre-calibrated for the corresponding material, the construction control unit will determine the corresponding overlap boundary as the overlap boundary for which rheological compensation needs to be performed, and adjust the material thickness, compaction frequency and single-point compaction duration of the next construction area near the overlap boundary according to the overlap compensation parameters corresponding to the current material type in the material performance parameter library.

[0123] In one implementation, when When the threshold is less than 0.85, the construction control unit performs rheological compensation at the lap joint interface. The threshold is predetermined based on the correspondence between interface state changes and construction parameters in the corresponding material's setting and hardening calibration record, and is stored in the material performance parameter library.

[0124] Within 150mm of the overlap boundary, the fabric execution unit increases the target fabric thickness by 8% to 12% from the original target fabric thickness. When the compaction execution unit moves to the corresponding position of the overlap boundary, it increases the vibration frequency and extends the single-point compaction duration based on the original compaction parameters.

[0125] The 150mm overlap compensation range, 8% to 12% fabric thickness adjustment, 15% vibration frequency adjustment, and 20% single-point compaction duration adjustment are pre-calibrated based on the corresponding material construction records and the allowable working range of the execution equipment, and stored in the material performance parameter library as overlap compensation parameters corresponding to the material type.

[0126] During the compaction compensation of the overlap interface, the displacement and pressure detection devices continue to collect the actual compaction response. The compensated material placement and compaction parameters are limited by the allowable operating range of the corresponding execution equipment. If the interface rheological attenuation coefficient does not meet the compensation conditions, construction of the next construction area is carried out according to the normal area construction parameters.

[0127] Step S111: Perform anomaly detection and safety procedures. The construction control system continuously monitors sensor status, material performance status, and construction equipment operating status. When the detected data exceeds the effective physical range of the sensor, the corresponding data is marked as invalid and will not be used for current construction quality prediction and parameter optimization.

[0128] If critical environmental or material performance conditions are unavailable, construction parameters should not be regenerated using incomplete input, and the equipment should be kept in a safe state. If the equipment reaches its permissible operating boundary but the actual construction response still fails to reach the target state, continue to increase control in the same adjustment direction.

[0129] Anomalies affecting equipment or personnel safety trigger the relevant actuators to stop; anomalies that allow construction to continue under restricted conditions limit the corresponding construction speed or execution parameters and output alarm information. Anomaly logs include the construction area identifier, occurrence time, anomaly data, and equipment operating status.

[0130] During the operation of the construction quality prediction model, the construction control unit continuously evaluates the degree of deviation of the current model input from the feature space of historical training samples.

[0131] The degree of deviation is determined based on the multivariate distance between the current standardized input and the statistical center of the historical training samples, and the covariance relationship between the input features of the historical training samples is used to correct the feature deviation in different directions, so that the correlation between the input features can participate in the judgment of the applicability range of the model.

[0132] In the model validation phase, a mapping relationship between distance and confidence level is established based on the construction quality prediction error under different feature spatial distances, and the confidence level boundaries corresponding to normal control state, restricted adjustment state and safety degradation state are determined.

[0133] During the on-site operation phase, the corresponding confidence index is obtained based on the current input feature space distance and the pre-established mapping relationship, eliminating the need to retrain the model during construction.

[0134] In one implementation, when the confidence index is greater than or equal to 0.70, the construction control unit evaluates the candidate mix design and determines the construction parameters according to the construction quality prediction results corresponding to the construction quality prediction model.

[0135] When the confidence index is greater than or equal to 0.40 and less than 0.70, the construction control unit continues to issue the construction mix ratio and initial construction parameters, and limits the single feedback adjustment range of the concrete placement unit, pouring unit and compaction unit relative to the current initial construction parameters to 50% of the adjustment range corresponding to the normal control state.

[0136] When the confidence index is less than 0.40, the construction control unit enters a safety downgrade state, stops the construction quality prediction model from automatically modifying the construction mix ratio for the next construction area, maintains the construction mix ratio that is currently being executed normally, and calls the refractory material rheological limit expert rule library to generate safety protection construction parameters.

[0137] The expert rule library for refractory material rheological limits stores the allowable ranges for flow state, placement state, pouring state, and compaction state for different material types. When entering a safety degradation state, the construction control unit indexes the corresponding rule entries according to the current material type and selects parameters that meet the allowable range of construction technology and equipment operating boundaries based on the current environmental conditions, forming safety protection construction parameters.

[0138] In one implementation, after entering the safety degradation state, the target pouring flow rate is reduced by 20% compared to before entering the safety degradation state, the compaction frequency is adjusted to the preset safe medium speed range of the corresponding compaction actuator, and an alarm message is output. The 20% pouring flow rate adjustment and the corresponding safe medium speed range are predetermined based on the current material construction process records and the calibration range of the actuator, and are stored in the refractory material rheological limit expert rule library.

[0139] During the safety downgrade period, each execution unit continues to implement local control within the allowable range of the equipment based on sensor feedback. The construction control unit continuously re-evaluates the model confidence level based on new field inputs; once the confidence level reaches the threshold corresponding to the normal control state, the process for determining construction parameters corresponding to the construction quality prediction model resumes.

[0140] Step S112: Establish the construction database and update the model. After the construction area is completed, write the environmental status data, material performance status data, actual construction mix proportions, concrete placement process data, pouring process data, compaction process data, and the available construction quality test results into the construction database.

[0141] Construction records for which quality inspection results have not yet been obtained are used for process traceability. Once the quality inspection results are obtained, they will be supplemented and linked to the corresponding construction area to form a complete historical construction record.

[0142] When updating the model, an updated dataset is constructed from historical construction records with complete input data and construction quality inspection results.

[0143] When updating the construction quality prediction model using transfer learning, the updated model is validated using historical construction records not involved in this update. If the prediction error of the updated model on the same validation records is no higher than that of the original construction quality prediction model, the current construction quality prediction model is replaced; otherwise, the original construction quality prediction model continues to be used.

[0144] Example 2: This example is based on Example 1 and is used for continuous large-area construction of the lining of a submerged arc furnace.

[0145] The continuous construction area is divided into multiple sub-areas arranged according to the construction sequence, and the construction control unit establishes an area identifier for each sub-area. Environmental status data, material performance status data, construction mix proportions, concrete placement process data, pouring process data, compaction process data, and abnormal data are all associated with the corresponding area identifiers.

[0146] Upon completion of the previous construction sub-area, a corresponding construction status evaluation result is generated. When construction begins in the next construction sub-area, environmental status data for the next construction sub-area is re-acquired, and the construction status evaluation result of the previous construction sub-area is used as feedback information to participate in the candidate mix design evaluation and construction parameter determination for the next construction sub-area.

[0147] When the environmental and material conditions of adjacent construction sub-regions are similar and the construction response of the previous region is stable, the actual construction parameters of the previous region are used as a reference for the candidate construction parameters of the subsequent region. When the environmental conditions change significantly or the material batches are changed, the candidate mix design evaluation is re-executed based on the actual environmental and material performance data obtained in the subsequent region.

[0148] Each construction equipment uses an independent local controller to complete the on-site adjustment of the actuator. The construction control unit is responsible for determining cross-process parameters, evaluating construction status, and providing feedback between regions, so as to maintain data continuity between cross-regional construction decisions and on-site equipment control.

[0149] The construction control unit includes an edge computing terminal and a cloud server. The edge computing terminal is deployed at the construction site of the submerged arc furnace and connects to the environmental condition sensing unit, material performance testing unit, concrete placement execution unit, pouring execution unit, and compaction execution unit. It is used to receive on-site data, call up local construction quality prediction models, generate construction control parameters, and execute on-site feedback control.

[0150] The cloud server stores the construction database and model training data, and updates the construction quality prediction model offline based on newly added construction records with construction quality inspection results. After the updated model is verified, a version of the model is released.

[0151] The edge computing terminal and the cloud server are connected via an industrial IoT gateway with data transmission resume capability after a communication interruption. When communication is interrupted, construction records that have not yet been uploaded are stored in the local cache of the edge computing terminal; after communication is restored, the records are uploaded again according to their timestamps and construction area identifiers.

[0152] The edge computing terminal locally stores the currently valid construction quality prediction model parameters, model data processing parameters, mix proportion data required for on-site construction, and construction control programs. When the communication network is interrupted, the edge computing terminal does not rely on the cloud server to return control results in real time, but continues to complete the mix proportion evaluation and construction parameter determination for the next construction area based on the on-site environmental conditions, material performance conditions, construction status evaluation results, and the locally cached construction quality prediction model.

[0153] After the updated model file is fully received and passes integrity verification, the model version switch is completed at the control boundary of the construction area to avoid the model transmission process affecting the on-site control being performed in the current construction area.

[0154] Example 3: This example is based on Example 1 and is used for situations where material type is changed or the construction environment is significantly different from the historical construction environment.

[0155] When the current material type differs from the material type corresponding to the main training data of the existing construction quality prediction model, historical records with similar material composition or construction performance are extracted from the construction database. If the current material has a feasible mix ratio corresponding to existing construction records, that mix ratio is used as a source of candidate mix ratios.

[0156] After obtaining new material construction records with construction quality inspection results, the corresponding construction records are added to the model update dataset, and the construction quality prediction model is updated using transfer learning. The updated model continues to use environmental state data, material performance state data, construction requirements, and candidate mix design features as model inputs.

[0157] When the historical construction records of new materials are insufficient to support model update and verification, continue to use the mix proportions and construction process parameters corresponding to the existing construction records, and collect on-site construction data; after obtaining sufficient construction quality test results, then perform model update and verification.

[0158] In high-temperature construction environments, the environmental condition sensing unit uses a detection device suitable for the corresponding environment to obtain environmental data, and uses the actual high-temperature environmental condition as input for candidate mix design evaluation and construction parameter determination.

[0159] Comparative Example 1:

[0160] This comparative example uses the same environmental condition acquisition, material performance testing, and construction equipment as Example 1. The difference is that after the environmental condition and material performance condition are used to determine the construction mix proportion, the material placement, pouring, and compaction adopt the pre-set process parameters. The monitoring results of each process are only used for the local control of the corresponding process and are not returned to the mix proportion and construction parameter determination process of the subsequent construction area.

[0161] This comparative example lacks a feedback path from the actual construction response in the current construction area to the construction decision-making in the next construction area. Example 1 uses the construction status evaluation results to evaluate candidate mix proportions and determine construction parameters for the next construction area, allowing the actual construction response in the previous area to participate in the construction decision-making in the next area.

[0162] Comparative Example 2:

[0163] This comparative example uses the same basic construction process as Example 1. The difference is that the construction quality prediction model only receives environmental state data and material performance state data, and does not use the mix composition information corresponding to the candidate mix as model input.

[0164] Under the same environmental and material performance conditions, different candidate mix proportions lack corresponding input differences, making it difficult to establish a data correspondence between changes in candidate mix proportions and construction quality prediction results. Example 1 inputs candidate mix proportion characteristics, environmental and material performance data into the construction quality prediction model, establishing a correspondence between the construction quality prediction results and specific candidate mix proportions.

[0165] To visually illustrate the differences in data transmission relationships between Example 1 and Comparative Examples 1 and 2, the comparison results are summarized below:

[0166] Table 1:

[0167] Comparison items Example 1 Comparative Example 1 Comparative Example 2 Candidate matching ratios participate in predicting inputs yes Can participate no The proportioning results are used to form process parameters. yes yes It is difficult to establish a definite correspondence between the predicted results and specific candidate proportions. The actual response of the three processes forms the regional construction status. yes no It can form Feedback of current area results to the next construction area yes no Feedback is welcome. A continuous data relationship is formed between the mix proportion prediction and on-site execution. yes no Incomplete

[0168] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0169] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. It should be noted that any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for controlling the construction of monolithic refractory materials for submerged arc furnaces using AI (Artificial Intelligence of Things) construction, characterized in that... include: Obtain environmental status data of the current construction area and perform performance testing on the unshaped refractory material to be constructed to obtain material performance status data; The environmental status data, material performance status data, construction requirements of the current construction area, and candidate mix characteristics corresponding to different candidate mix ratios are input into the construction quality prediction model to obtain the construction quality prediction results corresponding to each candidate mix ratio. Based on the construction quality prediction results corresponding to each candidate mix ratio, the construction mix ratio of the current construction area is determined, and the initial construction parameters for the material placement, pouring, and compaction processes are formed. The material placement, pouring, and compaction are carried out according to the initial construction parameters. The target compaction energy density of the compaction process is determined based on the material flow response of the pouring process and the actual material thickness of the material placement process. The construction parameters are adjusted according to the monitoring data of each process. By linking the monitoring data of the processes of material placement, pouring, and compaction, a construction status evaluation result for the current construction area is formed; The construction status evaluation results, along with the environmental status data and material performance data of the next construction area, are used to determine the construction parameters of the next construction area. The material placement and compaction parameters at the overlap position are then adjusted based on the overlap boundary temperature and the degree of hardening of the edge material. When the model input deviates from the feature space of historical training samples, the construction quality prediction model is restricted from modifying the construction mix ratio, and safety protection construction parameters are adopted; the process is executed cyclically according to the construction area sequence until the predetermined construction area is completed.

2. The AI ​​construction control method for monolithic refractory materials used in submerged arc furnaces according to claim 1, characterized in that, The environmental status data includes at least several data points, such as ambient temperature, ambient humidity, airflow status, and furnace shell temperature in the construction area. Data collected from different locations are filtered, calibrated, and time-aligned, and then an environmental status dataset is formed according to the corresponding construction area.

3. The AI ​​construction control method for monolithic refractory materials used in submerged arc furnaces according to claim 1, characterized in that, The material performance status data includes at least the chemical composition, moisture content, and flow properties of the unshaped refractory material; representative samples are obtained from the material to be constructed, and the material performance status data is obtained through composition detection, moisture content detection, and flow property detection, and the test results are correlated with the corresponding material batch and construction area.

4. The AI ​​construction control method for monolithic refractory materials used in submerged arc furnaces according to claim 1, characterized in that, The construction quality prediction model is established based on historical construction records, which include at least environmental conditions, material performance conditions, construction mix proportions, construction parameters, and post-construction quality inspection results. During training, the data corresponding to environmental conditions, material performance conditions, construction requirements, and construction mix proportions are used as model inputs, and the corresponding quality inspection results are used as supervisory information. The training objective is to reduce the prediction error between the construction quality prediction results and the corresponding quality inspection results, so that the construction quality prediction model can establish a mapping relationship between construction conditions, candidate mix proportions, and construction quality.

5. The AI ​​construction control method for monolithic refractory materials used in submerged arc furnaces according to claim 1, characterized in that, Candidate mix proportions that meet the material type and construction requirements are obtained from the mix proportion database, and the mix proportion composition information corresponding to each candidate mix proportion is used as the candidate mix proportion feature and input into the construction quality prediction model. The construction control unit forms the current input feature vector using the same data processing method as the training phase of the construction quality prediction model. Based on the covariance relationship between the input features of the historical training samples, it normalizes the multivariate deviation between the current input feature and the feature center of the historical training samples to obtain the distance evaluation result of the current input relative to the feature space of the historical training samples. The confidence index of the construction quality prediction result is determined based on the correspondence between the distance evaluation results and the construction quality prediction error described in the historical training phase. When the confidence index is lower than the preset safety threshold, the construction quality prediction model stops its autonomous modification of the construction mix ratio for the next construction area, maintains the construction mix ratio that is currently being executed normally, and calls the expert rule library of refractory material rheological limits to generate safety protection construction parameters.

6. The AI ​​construction control method for monolithic refractory materials for submerged arc furnaces according to claim 1, characterized in that, When placing materials, the target material placement state is determined based on the target construction thickness of the current construction area. The actual material thickness and actual material placement amount are obtained through the thickness detection device and material placement amount detection device set in the automatic material placement device. The feeding speed and the working state of the leveling mechanism are adjusted according to the deviation of the actual material thickness and actual material placement amount from the target material placement state. The adjusted material placement parameters and corresponding monitoring results are sent to the construction status evaluation process.

7. The AI ​​construction control method for monolithic refractory materials used in submerged arc furnaces according to claim 1, characterized in that, During the pouring process, the actual pouring volume flow rate is obtained through the flow detection unit, and the pipeline delivery pressure difference is obtained through the pressure detection units set at both ends of the pressure detection section. The driving state of the material delivery mechanism is adjusted according to the changes in the actual pouring volume flow rate and the pipeline delivery pressure difference relative to the corresponding target state. The actual pouring volume flow rate and the pipeline delivery pressure difference are then sent to the construction status evaluation process and the compaction parameter determination process.

8. The AI ​​construction control method for monolithic refractory materials for submerged arc furnaces according to claim 1, characterized in that, During the compaction process, based on the actual pouring volume flow rate monitored during the pouring process and the pipeline transport pressure difference at both ends of the pressure detection section, and combined with the pipeline structural parameters of the pressure detection section, the equivalent apparent viscosity used to characterize the actual transport resistance of the current material is determined. Then, combined with the actual material thickness obtained during the material placement process, the target compaction energy density of the current construction area is determined according to the following relationship: ; in, The target compaction energy density is expressed in J / m³. This is the energy density correction factor corresponding to the material type, in units of s⁻¹; The equivalent apparent viscosity is determined based on the actual pouring volume flow rate, pipeline delivery pressure difference, and pipeline structural parameters of the pressure testing section, and is expressed in Pa·s. This refers to the actual fabric thickness, in meters (m). The target construction thickness is expressed in meters (m). The thickness correction index corresponds to the material type and is dimensionless. The compaction execution unit is based on the target compaction energy density. The calibration relationship of the compaction equipment is adjusted to control the compaction force, vibration frequency, and single-point compaction duration. The displacement response and stress response of the material are obtained through displacement detection device and pressure detection device.

9. The AI ​​construction control method for monolithic refractory materials for submerged arc furnaces according to claim 1, characterized in that, When generating the construction status evaluation results, the thickness distribution state formed by the material placement process, the flow and pressure response state formed by the pouring process, and the displacement and stress response state formed by the compaction process are correlated according to the corresponding construction areas. The construction status evaluation results are determined based on the deviation of the actual response of each process from the target construction state of the current construction area.

10. An AI construction control system for monolithic refractory materials for submerged arc furnaces, used to execute the AI ​​construction control method for monolithic refractory materials for submerged arc furnaces as described in any one of claims 1 to 9, characterized in that, It includes an environmental condition sensing unit, a material performance testing unit, a construction control unit, a material placement execution unit, a pouring execution unit, and a compaction execution unit; The environmental condition sensing unit includes an environmental monitoring device arranged in the construction area of ​​the lining of the electric arc furnace, which is used to collect environmental condition data and send it to the construction control unit. The material performance testing unit includes a material sampling device and a testing device corresponding to the material performance parameters, which is used to obtain the material performance status data of the unshaped refractory material to be constructed and send it to the construction control unit. The construction control unit includes an edge computing terminal and a cloud server. The edge computing terminal is deployed at the construction site of the electric arc furnace and is connected to the environmental condition sensing unit, material performance testing unit, concrete placement execution unit, pouring execution unit, and compaction execution unit. It is used to call the local construction quality prediction model, generate construction control parameters, and receive process monitoring data. The cloud server is used to store the construction database and to offline train and update the construction quality prediction model based on new construction records. The material placement execution unit, pouring execution unit, and compaction execution unit each include a controller, a construction execution mechanism, and sensors for detecting the actual response of the corresponding process. The controller drives the construction execution mechanism according to the construction parameters issued by the edge computing terminal and feeds back the process monitoring data to the edge computing terminal. The edge computing terminal is connected to the cloud server through an industrial IoT gateway with data interruption and resume function; when the communication network is interrupted, the edge computing terminal independently completes the feedback control between construction areas based on the locally cached construction quality prediction model parameters and construction control data.