Steel liquid mold casting regulation system and method based on online oxygen concentration real-time monitoring

Abstract

The application discloses a molten steel mold casting regulation system and method based on online oxygen concentration real-time monitoring, relates to the technical field of molten steel mold casting, and realizes real-time collection of production data of a molten steel mold casting process based on a multi-source sensor arranged in a steel ingot mold; a digital twin model of the steel ingot mold is constructed based on model parameters of the steel ingot mold and the production data; during molten steel injection, a plurality of molten steel slices are generated based on time series and the digital twin model, mapping points are set on the molten steel slices, and predicted molten steel slices are mapped based on the mapping points; when oxygen activity data exceeds a preset multistage safety threshold, corresponding anti-oxidation operations are performed; when the anti-oxidation operations are performed, a search request is generated, the molten steel slices and the predicted molten steel slices at corresponding moments are connected in series based on an evolution path to generate a fault album; and early abnormal perception of the mold casting process and structured reuse of historical experience are realized through time series slicing and spatial interpolation of single-point oxygen activity data.

Description

Technical Field

[0001] This invention relates to the field of steel liquid mold casting technology, and specifically to a steel liquid mold casting control system and method based on real-time online oxygen concentration monitoring. Background Technology

[0002] In the existing technology, the oxygen activity control in the steel liquid mold casting process mainly relies on the online monitoring of single-point oxygen sensors arranged in the ladle or tundish. By collecting oxygen activity data in real time and comparing it with a preset threshold, the control operation is triggered when the oxygen activity exceeds the threshold.

[0003] However, existing technologies have certain shortcomings. Single-point oxygen sensors can only reflect the oxygen activity level near their installation location and cannot detect the spatial distribution differences of oxygen activity in different areas of molten steel within the ingot mold at the same time, resulting in a monitoring blind spot. At the same time, threshold-based control methods are only activated after oxygen activity exceeds the standard. By this time, irreversible oxide inclusion defects may have already formed in local areas of the molten steel. Furthermore, the correspondence between the operation time, oxygen activity deviation characteristics, control actions, and final effects of each anti-oxidation operation is not systematically recorded. Successful control experience cannot be directly reused in subsequent similar processes, and manual judgment is still required when the same type of oxygen activity abnormality occurs repeatedly.

[0004] Therefore, it is of great significance to develop a steel liquid mold casting control system based on real-time online oxygen concentration monitoring. Summary of the Invention

[0005] The purpose of this invention is to provide a steel liquid mold casting control system and method based on real-time online oxygen concentration monitoring, so as to solve the problems in the prior art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a steel liquid mold casting control system based on real-time online oxygen concentration monitoring, comprising:

[0007] Data acquisition module: Real-time acquisition of production data during the molten steel casting process based on multi-source sensors deployed inside the ingot mold;

[0008] Digital twin module: Connected to the data acquisition module, it constructs a digital twin model of the steel ingot mold based on the model parameters and production data.

[0009] Slice Management Module: Connected to the digital twin module, during the molten steel injection process, multiple molten steel slices are generated based on the time series and digital twin model. Mapping points are set on the molten steel slices, and predicted molten steel slices are mapped based on the mapping points.

[0010] Regulation and execution module: Connected to the slice management module, it executes the corresponding antioxidant operation when the oxygen activity data exceeds the preset safety threshold;

[0011] Fault Atlas Module: Connected to the control and execution module, when performing anti-oxidation operation, it generates a retrieval request and generates a fault atlas by concatenating the corresponding molten steel slices and predicted molten steel slices based on the evolution path.

[0012] In a preferred embodiment, the data acquisition module includes:

[0013] The oxygen sensing unit deploys multiple oxygen sensors at preset distances inside the steel ingot mold to collect oxygen activity data of molten steel in real time.

[0014] The operation data unit deploys multi-source sensors on the steel ingot mold to collect real-time operation data during steel molten injection. The operation data includes inert gas flow rate, inert gas concentration, injection speed, and wire feeding data.

[0015] In a preferred embodiment, the digital twin module includes:

[0016] The model building unit acquires the model parameters of the steel ingot mold and the deployment locations of multi-source sensors, and uses digital twin technology to construct a three-dimensional geometric model of the steel ingot mold;

[0017] The data mapping unit constructs a data transmission channel between the three-dimensional geometric model and the steel ingot mold, and receives production data in real time;

[0018] The three-dimensional geometric model is divided into multiple regions based on the deployment location of the oxygen sensing unit, and the normal oxygen activity benchmark for each region is preset.

[0019] The deoxidation evolution unit is used to construct the reaction model of molten steel. The reaction model includes the deoxidation reaction kinetic model and the inclusion formation model.

[0020] Based on the reaction model, a theoretical template for oxygen activity distribution is generated for each region. The theoretical template for oxygen activity distribution defines the relative proportion of oxygen activity distribution within the corresponding region.

[0021] In a preferred embodiment, the slice management module includes:

[0022] The slice generation unit is used to generate multiple molten steel slices based on a preset time interval during the molten steel injection process, and to mark the time interval for cutting each molten steel slice as a timestamp;

[0023] Each molten steel slice corresponds to a time interval in which molten steel is injected into the ingot mold;

[0024] The oxygen activity interpolation unit acquires oxygen activity data collected at corresponding time intervals, and uses an interpolation algorithm based on the relative proportion distribution relationship to interpolate the oxygen activity data onto the molten steel slice, generating an oxygen activity distribution map of the corresponding molten steel slice.

[0025] The mapping point unit compares the oxygen activity distribution map with the normal oxygen activity benchmark of the corresponding region of the molten steel slice for any given molten steel slice. Based on the comparison results, mapping points are generated, and oxygen activity evolution is predicted through the mapping points to generate predicted molten steel slices.

[0026] In a preferred embodiment, the step of generating mapping points based on the comparison results, and generating predicted molten steel slices using the mapping points, is as follows:

[0027] The preset allowable range and process allowable range are defined, and the allowable threshold is defined as the difference between the process allowable range and the normal oxygen activity benchmark.

[0028] For any molten steel slice, the oxygen activity distribution map of the molten steel slice is compared with the corresponding normal oxygen activity benchmark. If the difference falls within the allowable range, mapping points are deployed in the corresponding area of ​​the molten steel slice.

[0029] The mapping point calls the reaction model and generates a predicted molten steel slice based on the oxygen activity data of the corresponding region. The predicted molten steel slice includes a predicted timestamp, a predicted oxygen activity distribution map, an evolution path, and an estimated value of predicted inclusion generation.

[0030] Based on the iterative execution of oxygen activity evolution prediction based on predicted molten steel slices, a predicted molten steel slice tree with prediction timestamps as the hierarchy is generated.

[0031] Among them, the predicted molten steel slice tree takes the corresponding molten steel slice as the root node, the predicted molten steel slice as the intermediate node, and the evolution path as the edge;

[0032] Multiple molten steel slices correspond to the generated predicted molten steel slice trees, which together construct a predicted molten steel slice forest.

[0033] In a preferred embodiment, the control execution module includes:

[0034] The control unit presets multiple safety thresholds, including a first safety threshold, a second safety threshold, and a third safety threshold.

[0035] Real-time acquisition of oxygen activity data and oxygen activity distribution maps of molten steel slices; comparison of oxygen activity data and oxygen activity distribution maps with multi-level safety thresholds.

[0036] Antioxidant operations are generated based on the comparison results. The antioxidant operations include the operation time, oxygen activity data at the operation time, operation type, data correction amount, and operation result.

[0037] When the oxygen activity data exceeds the first safety threshold but does not exceed the second safety threshold, the first-level anti-oxidation operation is performed, which is to increase the inert gas flow rate or increase the inert gas concentration.

[0038] When the oxygen activity data exceeds the second safety threshold but does not exceed the third safety threshold, the second-level antioxidant operation is performed. The second-level antioxidant operation includes reducing the injection speed or increasing the wire feeding speed based on the first-level antioxidant operation.

[0039] When the oxygen activity data exceeds the third safety threshold, the third level of anti-oxidation operation is executed. The third level of anti-oxidation operation includes simultaneously increasing the inert gas flow rate, increasing the inert gas concentration, decreasing the injection speed, and increasing the wire feeding speed, and generating an alarm signal.

[0040] In a preferred embodiment, the fault map module includes:

[0041] The retrieval request unit generates a retrieval request when the digital twin model detects that the steel ingot mold has undergone an anti-oxidation operation.

[0042] The fault map generation unit obtains the operation time and oxygen activity data at the operation time based on the retrieval request, and searches for the predicted molten steel slice corresponding to the operation time in the predicted molten steel slice forest as the slice to be matched.

[0043] The oxygen activity data at the operation time is matched with the slice to be matched. If the matching degree exceeds the similarity threshold, the slice to be matched is marked as a similar predicted molten steel slice.

[0044] Tracing back along the evolution path from similar predicted molten steel slices to the root node molten steel slice, all molten steel slices and predicted molten steel slices on the reverse tracing path are connected in series according to the evolution path to generate a fault atlas, with the corresponding molten steel slice as the initial page;

[0045] The pre-adjustment instruction unit marks the operation type and data correction amount of the antioxidant operation on the corresponding mapping point on the initial page as a pre-adjustment instruction;

[0046] In the same type of casting process, when the deviation pattern of a certain mapping point matches the fault map, the pre-adjustment command is called to the control unit of the ingot mold, and the control unit preloads the data correction amount.

[0047] When a preset number of molten steel slices match the fault map, the corresponding preloaded data correction is performed.

[0048] This invention also provides a method for controlling steel liquid mold casting based on real-time online oxygen concentration monitoring, including:

[0049] Production data of the molten steel casting process is collected in real time by multi-source sensors deployed inside the steel ingot mold;

[0050] A digital twin model of the steel ingot mold is constructed based on the model parameters and production data.

[0051] During the molten steel injection process, multiple molten steel slices are generated based on time series and digital twin models. Mapping points are set on the molten steel slices, and predicted molten steel slices are mapped based on the mapping points.

[0052] When the oxygen activity data exceeds the preset multi-level safety threshold, the corresponding antioxidant operation is executed;

[0053] When performing the anti-oxidation operation, a retrieval request is generated, and the corresponding molten steel slices and predicted molten steel slices are connected in series based on the evolution path to generate a fault map.

[0054] The technical effects and advantages provided by the present invention in the above technical solution are as follows:

[0055] 1. This invention constructs a digital twin model and generates time-series steel melt slices, using real-time readings from a single-point oxygen sensor to reconstruct the spatial distribution map of oxygen activity for each slice. In traditional methods, a single sensor can only reflect changes in oxygen activity near its installation location and cannot detect differences in oxygen activity in different areas on the same cross section. This often results in blind spots where sensor readings are normal but local oxidation deterioration has occurred. This solution uses theoretical template calibration interpolation to expand single-point data into two-dimensional distribution information with spatial resolution, making the differences in oxygen activity in different areas such as the center, edge, and corner of the ingot mold explicit. The mapping point can accurately capture local anomalies even under a globally acceptable appearance, effectively improving the comprehensiveness and sensitivity of oxygen activity monitoring during the casting process.

[0056] 2. This invention, through the construction and reuse mechanism of the fault atlas, precipitates the deviation patterns, evolution paths, and data correction amounts corresponding to the anti-oxidation operation in a single casting process into a structured process knowledge atlas. When dealing with the same type of casting, the system matches the patterns in the fault atlas with real-time slices, and preloads the corresponding data correction amounts when the deviation characteristics of the mapping point match historical cases, thus transforming the control action from a post-event response to a pre-event deployment. Based on this mechanism, historical verification knowledge is transformed into knowledge assets that the system can autonomously call, effectively shortening the response chain for similar faults and improving the timeliness of control in the molten steel casting process. Attached Figure Description

[0057] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0058] Figure 1 This is a system flowchart of the present invention.

[0059] Figure 2 This is a flowchart of the method of the present invention. Detailed Implementation

[0060] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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, 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.

[0061] Example 1, please refer to Figure 1 As shown in this embodiment, the steel liquid mold casting control system based on real-time online oxygen concentration monitoring includes:

[0062] Data acquisition module: Real-time acquisition of production data during the molten steel casting process based on multi-source sensors deployed inside the ingot mold;

[0063] Digital twin module: Connected to the data acquisition module, it constructs a digital twin model of the steel ingot mold based on the model parameters and production data.

[0064] Slice Management Module: Connected to the digital twin module, during the molten steel injection process, multiple molten steel slices are generated based on the time series and digital twin model. Mapping points are set on the molten steel slices, and predicted molten steel slices are mapped based on the mapping points.

[0065] Regulation and execution module: Connected to the slice management module, it executes the corresponding antioxidant operation when the oxygen activity data exceeds the preset safety threshold;

[0066] Fault Atlas Module: Connected to the control and execution module, when an anti-oxidation operation is performed, a retrieval request is generated, and the corresponding molten steel slices and predicted molten steel slices are linked together based on the evolution path to generate a fault atlas.

[0067] Furthermore, in the prior art, the oxygen activity control in the steel liquid mold casting process mainly relies on the online monitoring of single-point oxygen sensors arranged in the ladle or tundish. By collecting oxygen activity data in real time and comparing it with a preset threshold, the control operation is triggered when the oxygen activity exceeds the threshold.

[0068] However, existing technologies have certain shortcomings. Single-point oxygen sensors can only reflect the oxygen activity level near their installation location and cannot detect the spatial distribution differences of oxygen activity in different areas of molten steel within the ingot mold at the same time, resulting in a monitoring blind spot. At the same time, threshold-based control methods are only activated after oxygen activity exceeds the standard. By this time, irreversible oxide inclusion defects may have already formed in local areas of the molten steel. Furthermore, the correspondence between the operation time, oxygen activity deviation characteristics, control actions, and final effects of each anti-oxidation operation is not systematically recorded. Successful control experience cannot be directly reused in subsequent similar processes, and manual judgment is still required when the same type of oxygen activity abnormality occurs repeatedly.

[0069] This invention constructs a digital twin model and generates time-series steel melt slices. It then uses real-time readings from a single-point oxygen sensor to reconstruct the spatial distribution map of oxygen activity for each slice. In traditional methods, a single sensor can only reflect changes in oxygen activity near its installation location and cannot detect differences in oxygen activity in different areas of the same cross-section. This often results in blind spots where sensor readings are normal but local oxidation deterioration has occurred. This solution uses theoretical template calibration interpolation to expand single-point data into two-dimensional distribution information with spatial resolution. This allows the differences in oxygen activity in different areas such as the center, edge, and corner of the ingot mold to be explicitly presented. The mapping points can accurately capture local anomalies even under the appearance of overall compliance, effectively improving the comprehensiveness and sensitivity of oxygen activity monitoring during the casting process.

[0070] By constructing and reusing fault maps, the deviation patterns, evolution paths, and data corrections corresponding to anti-oxidation operations in a single casting process are precipitated into a structured process knowledge map. When dealing with similar types of casting, the system matches patterns in the fault map with real-time slices, and preloads the corresponding data corrections when the deviation characteristics of the mapping point match historical cases, transforming control actions from post-event responses to pre-event deployments. Based on this mechanism, historical verification knowledge is transformed into knowledge assets that the system can autonomously call upon, effectively shortening the response chain for similar faults and improving the timeliness of control in the molten steel casting process.

[0071] In one embodiment, the data acquisition module includes:

[0072] The oxygen sensing unit deploys multiple oxygen sensors at preset distances inside the steel ingot mold to collect oxygen activity data of molten steel in real time.

[0073] The operation data unit deploys multi-source sensors on the steel ingot mold to collect real-time operation data during steel molten injection. The operation data includes inert gas flow rate, inert gas concentration, injection speed, and wire feeding data.

[0074] Furthermore, the oxygen sensing unit deploys multiple zirconia solid electrolyte oxygen sensors at preset distances along the height direction inside the ingot mold. Each oxygen sensor's front end penetrates the ingot mold wall to directly contact the molten steel, while its rear end integrates a temperature-compensated thermocouple. The millivolt-level electrical signal output by the oxygen sensor is transmitted via a shielded cable to the signal conditioning and analog-to-digital conversion subunit. This subunit filters, amplifies, and converts the signal to digital, then calculates and generates oxygen activity data in real time using the temperature compensation coefficient. The operation data unit deploys multi-source sensors on the ingot mold to collect operation data, including inert gas flow rate. The inert gas concentration is collected in real time by a mass flow meter installed on the protective argon pipeline, and the inert gas concentration is obtained by a gas analyzer arranged above the liquid surface of the ingot mold. The injection rate is calculated by combining the weight change rate output by the ladle weighing sensor with the cross-sectional area of ​​the ingot mold. The wire feeding data is recorded in real time by the rotary encoder on the wire feeder drive motor, which records the wire feeding length and speed. Furthermore, each oxygen activity data and operation data is marked with a unified timestamp and the corresponding liquid level height, forming a production data stream with spatiotemporal identification. The production data is then pushed to the digital twin module in real time through the industrial Ethernet interface.

[0075] In one embodiment, the digital twin module includes:

[0076] The model building unit acquires the model parameters of the steel ingot mold and the deployment locations of multi-source sensors, and uses digital twin technology to construct a three-dimensional geometric model of the steel ingot mold;

[0077] The data mapping unit constructs a data transmission channel between the three-dimensional geometric model and the steel ingot mold, and receives production data in real time;

[0078] The three-dimensional geometric model is divided into multiple regions based on the deployment location of the oxygen sensing unit, and the normal oxygen activity benchmark for each region is preset.

[0079] The deoxidation evolution unit is used to construct the reaction model of molten steel. The reaction model includes the deoxidation reaction kinetic model and the inclusion formation model.

[0080] Based on the reaction model, a theoretical template for oxygen activity distribution is generated for each region. The theoretical template for oxygen activity distribution defines the relative proportion of oxygen activity distribution within the corresponding region.

[0081] Furthermore, the model building unit imports the 3D scanned point cloud data of the ingot mold, extracts the internal dimensions, wall thickness, taper, and installation height and radial position parameters of the multi-source sensors, and uses a 3D modeling engine to generate a 3D geometric model of the ingot mold. This 3D geometric model includes the installation coordinates of each oxygen sensor and the spatial position of the inert gas pipeline outlet. The data mapping unit establishes a data transmission channel with the data acquisition module via the industrial Ethernet protocol to receive production data streams in real time. Based on the installation coordinates of each oxygen sensor in the oxygen sensing unit, the 3D geometric model is divided into multiple regions along the liquid surface rise direction, corresponding one-to-one with each oxygen sensor. Each region extends along the liquid surface rise direction, covering a preset height range, with the installation height of the corresponding oxygen sensor as the reference. The data mapping unit presets a set of normal oxygen activity benchmarks for each region. These benchmarks define the oxygen activity of the molten steel within the corresponding region under ideal deoxidation and ideal protective atmosphere conditions at different time periods and radial positions. The value range; the reaction model constructed by the deoxidation evolution unit includes a deoxidation reaction kinetic model and an inclusion formation model. The deoxidation reaction kinetic model describes the relationship between the reaction rate of deoxidizing elements and dissolved oxygen and the temperature of the molten steel based on the Arrhenius equation. The inclusion formation model describes the nucleation rate, growth rate of deoxidation products, and the collision aggregation and floating behavior of inclusions in molten steel based on the classical nucleation and growth theory. Based on the reaction model, a theoretical template for the oxygen activity distribution corresponding to each region is generated. The steps are as follows: taking the corresponding region as the computational domain, taking the initial composition of the steel grade and the casting temperature as the initial conditions, and taking the inert gas flow rate and inert gas concentration as the boundary conditions, inputting them into the reaction model, obtaining the spatial distribution of oxygen activity in the radial section in this region through steady-state solution, extracting the oxygen activity calculation value at the oxygen sensor installation coordinate position as the reference value, and performing a ratio calculation between the oxygen activity calculation value of each other point on the section and the reference value to obtain the relative proportionality coefficient of each point. All relative proportionality coefficients constitute the theoretical template for the oxygen activity distribution in this region.

[0082] In one embodiment, the slice management module includes:

[0083] The slice generation unit is used to generate multiple molten steel slices based on a preset time interval during the molten steel injection process, and to mark the time interval for cutting each molten steel slice as a timestamp;

[0084] Each molten steel slice corresponds to a time interval in which molten steel is injected into the ingot mold;

[0085] The oxygen activity interpolation unit acquires oxygen activity data collected at corresponding time intervals, and uses an interpolation algorithm based on the relative proportion distribution relationship to interpolate the oxygen activity data onto the molten steel slice, generating an oxygen activity distribution map of the corresponding molten steel slice.

[0086] The mapping point unit compares the oxygen activity distribution map with the normal oxygen activity benchmark of the corresponding area of ​​the molten steel slice for any given molten steel slice. Based on the comparison results, mapping points are generated, and oxygen activity evolution is predicted through the mapping points to generate predicted molten steel slices.

[0087] Furthermore, during the molten steel injection process, the slice generation unit periodically triggers slice generation operations at preset time intervals, with the ladle opening time as the zero point. Each molten steel slice is assigned a timestamp to mark the start and end points of its corresponding time interval, while simultaneously recording the range of liquid level height in the ingot mold within that time interval. The oxygen activity interpolation unit acquires the oxygen activity data sequence collected by the oxygen sensing unit within that time interval, calculates the time-weighted average of the oxygen activity data within that time interval as the calibration point oxygen activity value for the corresponding molten steel slice, retrieves the oxygen activity distribution theoretical template for the region to which the height interval of the molten steel slice belongs from the digital twin module, and divides the calibration point oxygen activity value by the relative proportionality coefficient of the calibration point position in the oxygen activity distribution theoretical template to obtain the calibration factor. The calibration factor is multiplied by the relative scaling factor of the remaining grid nodes in the oxygen activity distribution theoretical template to obtain the oxygen activity interpolation of each grid node. The oxygen activity interpolation of all grid nodes together constitutes the oxygen activity distribution map of the molten steel slice. For any molten steel slice, the mapping point unit compares the oxygen activity distribution map of the molten steel slice with the normal oxygen activity benchmark of the corresponding area in the digital twin module, grid by grid node. When the oxygen activity distribution map value of a certain grid node does not exceed the process allowable range but deviates from the corresponding value range of the grid node in the normal oxygen activity benchmark, the grid node falls within the allowable range. A mapping point is deployed at the location of the grid node. The mapping point records the spatial coordinates of the grid node, the current oxygen activity value, and the value relative to the normal oxygen activity benchmark. The deviation of the oxygen activity baseline; the deployment of mapping points activates the oxygen activity evolution prediction process. The mapping point calls the reaction model in the deoxidation evolution unit of the digital twin module, using the oxygen activity value and deviation of the grid node where the mapping point is located as the initial state, and the current running data as the boundary conditions, to extrapolate to multiple preset prediction time steps. Each prediction time step corresponds to the generation of a predicted molten steel slice. The predicted molten steel slice includes a prediction timestamp, a predicted oxygen activity distribution map, an evolution path, and an estimated value of predicted inclusion generation. The evolution path records which mapping point and which extrapolation branch generated the predicted molten steel slice. Based on the current predicted molten steel slice, the oxygen activity evolution prediction is iteratively executed, that is, each predicted molten steel slice is used as... The new input calls the reaction model again to generate the predicted molten steel slice for the next prediction time step, until the predicted oxygen activity distribution map returns to the normal oxygen activity benchmark or a predicted molten steel slice shows an oxygen activity interpolation exceeding the multi-level safety threshold. All the predicted molten steel slices generated in the iteration form a predicted molten steel slice tree with the molten steel slice where the initial mapping point is located as the root node according to the generation level and parent-child relationship. Each node in the predicted molten steel slice tree is a molten steel slice or a predicted molten steel slice. The nodes are connected by evolutionary edges, which carry the operating data conditions on which the deduction is based in this step. Multiple mapping points deployed on multiple molten steel slices generate their own corresponding predicted molten steel slice trees. All the predicted molten steel slice trees together construct the predicted molten steel slice forest for the current casting process.

[0088] In one embodiment, the step of generating mapping points based on the comparison results, and generating predicted molten steel slices using the mapping points, is as follows:

[0089] The preset allowable range and process allowable range are defined, and the allowable threshold is defined as the difference between the process allowable range and the normal oxygen activity benchmark.

[0090] For any molten steel slice, the oxygen activity distribution map of the molten steel slice is compared with the corresponding normal oxygen activity benchmark. If the difference falls within the allowable range, mapping points are deployed in the corresponding area of ​​the molten steel slice.

[0091] The mapping point calls the reaction model and generates a predicted molten steel slice based on the oxygen activity data of the corresponding region. The predicted molten steel slice includes a predicted timestamp, a predicted oxygen activity distribution map, an evolution path, and an estimated value of predicted inclusion generation.

[0092] Based on the iterative execution of oxygen activity evolution prediction based on predicted molten steel slices, a predicted molten steel slice tree with prediction timestamps as the hierarchy is generated.

[0093] Among them, the predicted molten steel slice tree takes the corresponding molten steel slice as the root node, the predicted molten steel slice as the intermediate node, and the evolution path as the edge;

[0094] Multiple molten steel slices correspond to the generated predicted molten steel slice trees, which together construct a predicted molten steel slice forest;

[0095] Furthermore, the permissible range of the process is determined based on the upper and lower limits of oxygen activity specified in the casting process specification for this steel grade. Exceeding the upper limit of oxygen activity triggers an anti-oxidation operation. The normal oxygen activity benchmark is pre-stored for each region by the data mapping unit of the digital twin module. Its value comes from the standard distribution band obtained by statistical regression of the oxygen activity distribution data of the same region and time of the historical best heat. The allowable interval is defined as the difference between the set of permissible process ranges and the set of normal oxygen activity benchmarks, that is, the interval consisting of all oxygen activity values ​​that belong to the permissible process range but not to the normal oxygen activity benchmark. For any steel slice, the mapping point unit maps each grid node in the oxygen activity distribution map of the steel slice. The oxygen activity interpolation is compared with the standard oxygen activity value corresponding to the normal oxygen activity benchmark for that grid node. If the difference falls within the allowable range, a mapping point is deployed at the spatial coordinates of that grid node. The mapping point records the spatial coordinates of the grid node, the current oxygen activity value, the deviation from the normal oxygen activity benchmark, and the activation timestamp of the mapping point. After the mapping point is deployed, the reaction model in the deoxygenation evolution unit of the digital twin module is immediately invoked. The oxygen activity value and deviation of the grid node where the mapping point is located are used as the initial state input, and the current running data is used as the boundary conditions. The model is then extrapolated and solved for multiple preset prediction time steps. Each prediction time step generates a prediction. The predicted molten steel slice includes a predicted timestamp, a predicted oxygen activity distribution map, an evolution path, and a predicted inclusion formation estimate. The predicted timestamp is the future time interval corresponding to the predicted molten steel slice. The predicted oxygen activity distribution map is the spatial distribution of oxygen activity within the predicted time step obtained from the reaction model. The evolution path records the complete inference branch sequence from the root node molten steel slice to the predicted molten steel slice. The predicted inclusion formation estimate is the quantity estimate and spatial distribution of newly formed inclusions within the predicted time step obtained from the inclusion formation model. The specific implementation method of iteratively performing oxygen activity evolution prediction based on the predicted molten steel slice is as follows: each predicted molten steel slice is used as the initial state for a new round of inference. The system takes a state input and calls the reaction model again to generate a predicted molten steel slice for the next prediction time step. The iteration process continues until the predicted oxygen activity distribution map returns to the normal oxygen activity baseline range or a predicted molten steel slice shows an oxygen activity interpolation exceeding a multi-level safety threshold. The predicted molten steel slices generated iteratively are arranged according to the generation hierarchy. The first layer of predicted molten steel slices takes the root node molten steel slice as its parent node, the second layer of predicted molten steel slices takes the first layer of predicted molten steel slices as its parent node, and so on to form a predicted molten steel slice tree. The root node of the predicted molten steel slice tree is the molten steel slice of the deployment mapping point, the intermediate nodes are the predicted molten steel slices of each layer, and the nodes are connected by evolutionary edges, which carry the operating data conditions on which the deduction is based in this step.In the current casting process, each mapping point deployed on all molten steel slices generates a corresponding predicted molten steel slice tree. All predicted molten steel slice trees are indexed in the digital twin model according to the timestamp and spatial location of their respective root node molten steel slices, collectively constructing a predicted molten steel slice forest covering the entire current casting process.

[0096] In one embodiment, the control execution module includes:

[0097] The control unit presets multiple safety thresholds, including a first safety threshold, a second safety threshold, and a third safety threshold.

[0098] Real-time acquisition of oxygen activity data and oxygen activity distribution maps of molten steel slices; comparison of oxygen activity data and oxygen activity distribution maps with multi-level safety thresholds.

[0099] Antioxidant operations are generated based on the comparison results. The antioxidant operations include the operation time, oxygen activity data at the operation time, operation type, data correction amount, and operation result.

[0100] When the oxygen activity data exceeds the first safety threshold but does not exceed the second safety threshold, the first-level anti-oxidation operation is performed, which is to increase the inert gas flow rate or increase the inert gas concentration.

[0101] When the oxygen activity data exceeds the second safety threshold but does not exceed the third safety threshold, the second-level antioxidant operation is performed. The second-level antioxidant operation includes reducing the injection speed or increasing the wire feeding speed based on the first-level antioxidant operation.

[0102] When the oxygen activity data exceeds the third safety threshold, the third level of anti-oxidation operation is executed. The third level of anti-oxidation operation includes simultaneously increasing the inert gas flow rate, increasing the inert gas concentration, decreasing the injection speed, and increasing the wire feeding speed, and generating an alarm signal.

[0103] Furthermore, based on the process specifications and historical production data of this steel grade, multiple gradient nodes are determined for the upper limit of the oxygen activity safety threshold. The first safety threshold is set as the upper limit of the process-allowed range multiplied by a first coefficient; the second safety threshold is set as the upper limit of the process-allowed range multiplied by a second coefficient; and the third safety threshold is set as the upper limit of the process-allowed range multiplied by a third coefficient. The first coefficient is less than the second coefficient, and the second coefficient is less than the third coefficient, thus forming a progressively increasing safety threshold sequence. During the casting process, the control unit continuously receives the oxygen activity data stream output by the oxygen sensing unit. The oxygen activity distribution map of the latest steel melt slice generated by the slice management module is used to compare the current oxygen activity value in the oxygen activity data stream with the multi-level safety thresholds step by step. Simultaneously, the oxygen activity value of each grid node in the oxygen activity distribution map is compared node by node with the multi-level safety thresholds. Based on the comparison results, an anti-oxidation operation is generated. The anti-oxidation operation includes the operation time, the oxygen activity data at the operation time, the operation type, the data correction amount, and the operation result. The operation time is defined as the system clock reading when the control unit first detects that the oxygen activity data exceeds a certain level of safety threshold, and the oxygen activity data at the operation time is the oxygen activity value output by the oxygen sensing unit at that time. The values ​​and corresponding oxygen activity distribution maps; the operation type is divided into Level 1, Level 2, or Level 3 antioxidant operation based on the level of safety threshold exceeded; the data correction amount is the operating data adjustment value calculated by the control unit based on the deviation between the current oxygen activity data and the corresponding safety threshold. When performing Level 1 antioxidant operation, the data correction amount includes the inert gas flow rate increment or the inert gas concentration increment, and the control unit sends a correction command to the flow regulating valve or gas proportioning valve of the inert gas pipeline; when performing Level 2 antioxidant operation, the data correction amount is based on Level 1 antioxidant operation, plus the injection rate reduction or... The feed rate increment is achieved by the control unit sending a command to decrease the injection rate to the ladle sliding nozzle actuator or a command to increase the feed rate to the wire feeder controller. When performing the third-level anti-oxidation operation, the data correction includes the inert gas flow rate increment, inert gas concentration increment, injection rate reduction, and feed rate increment. The control unit sends correction commands to all corresponding actuators in parallel and sends an alarm signal to the operation terminal. The operation result is the system clock reading when the oxygen activity data falls below the first safety threshold after the anti-oxidation operation is performed, and the oxygen activity value after the fall. This operation result is recorded and used for the effect evaluation of the anti-oxidation operation.

[0104] In one embodiment, the fault map module includes:

[0105] The retrieval request unit generates a retrieval request when the digital twin model detects that the steel ingot mold has undergone an anti-oxidation operation.

[0106] The fault map generation unit obtains the operation time and oxygen activity data at the operation time based on the retrieval request, and searches for the predicted molten steel slice corresponding to the operation time in the predicted molten steel slice forest as the slice to be matched.

[0107] The oxygen activity data at the operation time is matched with the slice to be matched. If the matching degree exceeds the similarity threshold, the slice to be matched is marked as a similar predicted molten steel slice.

[0108] Tracing back along the evolution path from similar predicted molten steel slices to the root node molten steel slice, all molten steel slices and predicted molten steel slices on the reverse tracing path are connected in series according to the evolution path to generate a fault atlas, with the corresponding molten steel slice as the initial page;

[0109] The pre-adjustment instruction unit marks the operation type and data correction amount of the antioxidant operation on the corresponding mapping point on the initial page as a pre-adjustment instruction;

[0110] In the same type of casting process, when the deviation pattern of a certain mapping point matches the fault map, the pre-adjustment command is called to the control unit of the ingot mold, and the control unit preloads the data correction amount.

[0111] When a preset number of molten steel slices match the fault map, the corresponding preloaded data correction is executed;

[0112] Furthermore, the retrieval request unit detects the execution of the antioxidant operation by monitoring changes in the antioxidant operation execution status flag. When the control unit completes an antioxidant operation and writes the operation record to the operation log, the status flag is set. Upon capturing this set signal, the retrieval request unit immediately generates a retrieval request containing the operation time and oxygen activity data at that time. After receiving the retrieval request, the fault map generation unit extracts the operation time as the index key, traverses all nodes of all predicted molten steel slice trees in the predicted molten steel slice forest, and filters out predicted molten steel slices whose time deviation between the predicted timestamp and the operation time is within a preset time tolerance range, as the set of slices to be matched. The oxygen activity data at the operation time is then used to... The similarity is calculated based on the predicted oxygen activity distribution map of each slice in the set of slices to be matched. The similarity calculation adopts the spatial distribution correlation coefficient method, comparing the numerical deviation between the oxygen activity distribution map and the predicted oxygen activity distribution map at each grid node. When the correlation coefficient exceeds the preset similarity threshold, the slice to be matched is marked as a similar predicted molten steel slice. The fault map generation unit starts from the similar predicted molten steel slice and recursively traces back along the evolution edge, visiting the parent node of the similar predicted molten steel slice in turn until the root node molten steel slice is reached. All molten steel slices and predicted molten steel slices passed in turn along the reverse tracing path are connected in the forward order of the evolution path. Each slice uses its timestamp and oxygen activity distribution map as page content to generate a... A complete fault evolution atlas is generated, with the first page of the atlas being the root node molten steel slice representing the starting point of the tracing path. The pre-adjustment instruction unit extracts the operation type and data correction amount of the current anti-oxidation operation, locates the mapping point that initially triggered the evolution path on the first page of the fault atlas, i.e., the root node molten steel slice, and marks the operation type and data correction amount in the attribute set of that mapping point as a pre-adjustment instruction. In the subsequent casting process of ingot molds of the same type and specification of steel, when the slice management module generates molten steel slices and deploys mapping points, it matches the deviation pattern of each newly deployed mapping point with the deviation patterns of all first page mapping points in the stored fault atlas set. The deviation pattern matching uses multi-dimensional feature vector similarity calculation, where the feature vector contains the spatial coordinates of the mapping points. The system identifies the target, deviation amount, and local oxygen activity gradient direction. When the deviation pattern of a certain mapping point successfully matches the deviation pattern of a mapping point on the first page of a fault map, the pre-adjustment instruction marked in the fault map is sent to the control unit. The control unit loads the data correction amount in the pre-adjustment instruction into a ready state. As the casting process progresses, newly generated molten steel slices are sequentially matched with slices on subsequent pages of the fault map. When the number of consecutively matched molten steel slices reaches a preset matching quantity threshold, the control unit sends the data correction amount in the ready state to each actuator to perform preventive control in advance. If the matching chain is interrupted before reaching the preset matching quantity threshold, the control unit cancels the pre-loaded data correction amount, and the current pre-adjustment process terminates.

[0113] Example 2, please refer to Figure 2 As shown in this embodiment, the steel liquid mold casting control method based on real-time online oxygen concentration monitoring includes:

[0114] Production data of the molten steel casting process is collected in real time by multi-source sensors deployed inside the steel ingot mold;

[0115] A digital twin model of the steel ingot mold is constructed based on the model parameters and production data.

[0116] During the molten steel injection process, multiple molten steel slices are generated based on time series and digital twin models. Mapping points are set on the molten steel slices, and predicted molten steel slices are mapped based on the mapping points.

[0117] When the oxygen activity data exceeds the preset multi-level safety threshold, the corresponding antioxidant operation is executed;

[0118] When performing the anti-oxidation operation, a retrieval request is generated, and the corresponding molten steel slices and predicted molten steel slices are connected in series based on the evolution path to generate a fault map.

[0119] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A steel liquid mold casting control system based on real-time online oxygen concentration monitoring, characterized in that, include: Data acquisition module: Real-time acquisition of production data during the molten steel casting process based on multi-source sensors deployed inside the ingot mold; The data acquisition module includes an oxygen sensing unit, in which multiple oxygen sensors are deployed at a preset distance inside the steel ingot mold to collect oxygen activity data of molten steel in real time. The operation data unit deploys multi-source sensors on the steel ingot mold to collect real-time operation data during steel molten injection. The operation data includes inert gas flow rate, inert gas concentration, injection speed, and wire feeding data. Digital twin module: Connected to the data acquisition module, it constructs a digital twin model of the steel ingot mold based on the model parameters and production data. Slice Management Module: Connected to the digital twin module, during the molten steel injection process, multiple molten steel slices are generated based on the time series and digital twin model. Mapping points are set on the molten steel slices, and predicted molten steel slices are mapped based on the mapping points. The slice management module includes: a slice generation unit, used to generate multiple molten steel slices based on a preset time interval during the molten steel injection process, and to mark the time interval for cutting each molten steel slice as a timestamp; Each molten steel slice corresponds to a time interval in which molten steel is injected into the ingot mold; The oxygen activity interpolation unit acquires oxygen activity data collected at corresponding time intervals, and uses an interpolation algorithm based on the relative proportion distribution relationship to interpolate the oxygen activity data onto the molten steel slice, generating an oxygen activity distribution map of the corresponding molten steel slice. The mapping point unit compares the oxygen activity distribution map with the normal oxygen activity benchmark of the corresponding area of ​​the molten steel slice for any given molten steel slice. Based on the comparison results, mapping points are generated, and oxygen activity evolution is predicted through the mapping points to generate predicted molten steel slices. Regulation and execution module: Connected to the slice management module, it executes the corresponding antioxidant operation when the oxygen activity data exceeds the preset safety threshold; Fault Atlas Module: Connected to the control and execution module, when performing anti-oxidation operation, it generates a retrieval request and generates a fault atlas by concatenating the corresponding molten steel slices and predicted molten steel slices based on the evolution path.

2. The steel liquid mold casting control system based on real-time online oxygen concentration monitoring according to claim 1, characterized in that, The digital twin module includes: The model building unit acquires the model parameters of the steel ingot mold and the deployment locations of multi-source sensors, and uses digital twin technology to construct a three-dimensional geometric model of the steel ingot mold; The data mapping unit constructs a data transmission channel between the three-dimensional geometric model and the steel ingot mold, and receives production data in real time; The three-dimensional geometric model is divided into multiple regions based on the deployment location of the oxygen sensing unit, and the normal oxygen activity benchmark for each region is preset. The deoxidation evolution unit is used to construct the reaction model of molten steel. The reaction model includes the deoxidation reaction kinetic model and the inclusion formation model. Based on the reaction model, a theoretical template for oxygen activity distribution is generated for each region. The theoretical template for oxygen activity distribution defines the relative proportion of oxygen activity distribution within the corresponding region.

3. The steel liquid mold casting control system based on real-time online oxygen concentration monitoring according to claim 1, characterized in that, The steps for generating mapping points based on the comparison results, and generating predicted molten steel slices using these mapping points, are as follows: The preset allowable range and process allowable range are defined, and the allowable threshold is defined as the difference between the process allowable range and the normal oxygen activity benchmark. For any molten steel slice, the oxygen activity distribution map of the molten steel slice is compared with the corresponding normal oxygen activity benchmark. If the difference falls within the allowable range, mapping points are deployed in the corresponding area of ​​the molten steel slice. The mapping point calls the reaction model and generates a predicted molten steel slice based on the oxygen activity data of the corresponding region. The predicted molten steel slice includes a predicted timestamp, a predicted oxygen activity distribution map, an evolution path, and an estimated value of predicted inclusion generation. Based on the iterative execution of oxygen activity evolution prediction based on predicted molten steel slices, a predicted molten steel slice tree with prediction timestamps as the hierarchy is generated. Among them, the predicted molten steel slice tree takes the corresponding molten steel slice as the root node, the predicted molten steel slice as the intermediate node, and the evolution path as the edge; Multiple molten steel slices correspond to the generated predicted molten steel slice trees, which together construct a predicted molten steel slice forest.

4. The steel liquid mold casting control system based on real-time online oxygen concentration monitoring according to claim 1, characterized in that, The control execution module includes: The control unit presets multiple safety thresholds, including a first safety threshold, a second safety threshold, and a third safety threshold. Real-time acquisition of oxygen activity data and oxygen activity distribution maps of molten steel slices; comparison of oxygen activity data and oxygen activity distribution maps with multi-level safety thresholds. Antioxidant operations are generated based on the comparison results. The antioxidant operations include the operation time, oxygen activity data at the operation time, operation type, data correction amount, and operation result. When the oxygen activity data exceeds the first safety threshold but does not exceed the second safety threshold, the first-level anti-oxidation operation is performed, which is to increase the inert gas flow rate or increase the inert gas concentration. When the oxygen activity data exceeds the second safety threshold but does not exceed the third safety threshold, the second-level antioxidant operation is performed. The second-level antioxidant operation includes reducing the injection speed or increasing the wire feeding speed based on the first-level antioxidant operation. When the oxygen activity data exceeds the third safety threshold, the third level of anti-oxidation operation is executed. The third level of anti-oxidation operation includes simultaneously increasing the inert gas flow rate, increasing the inert gas concentration, decreasing the injection speed, and increasing the wire feeding speed, and generating an alarm signal.

5. The steel liquid mold casting control system based on real-time online oxygen concentration monitoring according to claim 4, characterized in that, The fault map module includes: The retrieval request unit generates a retrieval request when the digital twin model detects that the steel ingot mold has undergone an anti-oxidation operation. The fault map generation unit obtains the operation time and oxygen activity data at the operation time based on the retrieval request, and searches for the predicted molten steel slice corresponding to the operation time in the predicted molten steel slice forest as the slice to be matched. The oxygen activity data at the operation time is matched with the slice to be matched. If the matching degree exceeds the similarity threshold, the slice to be matched is marked as a similar predicted molten steel slice. Tracing back along the evolution path from similar predicted molten steel slices to the root node molten steel slice, all molten steel slices and predicted molten steel slices on the reverse tracing path are connected in series according to the evolution path to generate a fault atlas, with the corresponding molten steel slice as the initial page; The pre-adjustment instruction unit marks the operation type and data correction amount of the antioxidant operation on the corresponding mapping point on the initial page as a pre-adjustment instruction; In the same type of casting process, when the deviation pattern of a certain mapping point matches the fault map, the pre-adjustment command is called to the control unit of the ingot mold, and the control unit preloads the data correction amount. When a preset number of molten steel slices match the fault map, the corresponding preloaded data correction is performed.

6. A method for controlling steel liquid mold casting based on real-time online oxygen concentration monitoring, used to implement the steel liquid mold casting control system based on real-time online oxygen concentration monitoring as described in any one of claims 1-5, characterized in that, Production data of the molten steel casting process is collected in real time by multi-source sensors deployed inside the steel ingot mold; A digital twin model of the steel ingot mold is constructed based on the model parameters and production data. During the molten steel injection process, multiple molten steel slices are generated based on time series and digital twin models. Mapping points are set on the molten steel slices, and predicted molten steel slices are mapped based on the mapping points. When the oxygen activity data exceeds the preset multi-level safety threshold, the corresponding antioxidant operation is executed; When performing the anti-oxidation operation, a retrieval request is generated, and the corresponding molten steel slices and predicted molten steel slices are connected in series based on the evolution path to generate a fault map.

Images