Method, system, equipment and medium for determining process nodes of inclusions in continuous casting slabs

Abstract

The present disclosure provides a method, system, device, and medium for determining process nodes for inclusions in continuous casting billets. The determination method includes: obtaining sample data of several billets during the continuous casting process; obtaining a decision tree model based on the sample data; modifying the decision tree model based on preset standards; and obtaining a target process node corresponding to the inclusions in the continuous casting billet based on the modified decision tree model. The present disclosure obtains a decision tree model through sample data of several billets during the continuous casting process, and then modifies the decision tree model based on preset standards to obtain a target process node corresponding to the inclusions in the continuous casting billet based on the modified decision tree model. This conveniently combines preset standards such as expert judgment experience with root cause judgment rules mined from data, thereby locating the root cause of inclusions and accurately identifying the root cause of the inclusion problem. The method is suitable for scenarios with large amounts of data, complex problems, and the need to be combined with the background of the steel production industry, and has high credibility.

Description

Technical Field

[0001] The present disclosure relates to the technical field of steel production, and in particular to a method, system, equipment and medium for determining process nodes of inclusions in continuous casting billets. Background Art

[0002] Currently, during the continuous casting process of steel production, impurities or inclusions appear in the cast billet. Common inclusions include oxides, sulfides, nitrides, and other non-metallic particles.

[0003] Identifying the root cause of inclusion problems in billets is essential so that effective measures can be taken to resolve the issue and prevent its recurrence. This is known as root cause location. Root cause location is often used in areas such as quality management, troubleshooting, and continuous improvement to improve the quality and reliability of steel production.

[0004] Then, during the continuous casting process, the amount of data is large and there are multiple factors that lead to inclusion problems, making root cause location more complicated and unable to accurately identify the root cause of the inclusion problem. Summary of the Invention

[0005] The technical problem to be solved by the present disclosure is to overcome the defects in the prior art such as the large amount of data in the continuous casting process, the presence of multiple factors leading to inclusion problems, the complexity of root cause location, and the inability to accurately identify the root cause of the inclusion problem, and to provide a method, system, equipment and medium for determining the process nodes of inclusions in continuous casting billets.

[0006] The present disclosure solves the above technical problems through the following technical solutions:

[0007] The present disclosure provides a method for determining a process node of inclusions in a continuous casting slab, the method comprising:

[0008] Obtain sample data of several billets during the continuous casting process;

[0009] The sample data includes process data and corresponding inclusion status data, and the process data includes a number of initial process nodes and corresponding actual values of the nodes;

[0010] Based on the sample data, a decision tree model is obtained;

[0011] The decision tree model includes leaf nodes and non-leaf nodes, the leaf nodes include the inclusion state data, and the non-leaf nodes include the initial process nodes and corresponding node thresholds;

[0012] Modifying the decision tree model based on preset criteria;

[0013] Based on the modified decision tree model, the target process node corresponding to the inclusion of the continuous casting billet is obtained.

[0014] Optionally, in response to the preset standard including historical experience data of the continuous casting process, the step of modifying the decision tree model based on the preset standard includes:

[0015] Based on the historical experience data, modify the node threshold of the non-leaf node in the decision tree model;

[0016] and / or,

[0017] Based on the historical experience data, modifying the initial process nodes of the non-leaf nodes in the decision tree model; and modifying the decision tree model based on the modified initial process nodes;

[0018] and / or,

[0019] Based on the historical experience data, deleting some decision paths of the decision tree model and / or some subtrees of the decision tree model;

[0020] and / or,

[0021] Adding a decision path from the root node of the decision tree model based on the historical experience data;

[0022] and / or,

[0023] In response to the preset criteria including the process principle of the continuous casting process, the step of modifying the decision tree model based on the preset criteria includes:

[0024] Determining whether the node threshold in the decision path of the decision tree model meets the requirements of the process principle;

[0025] In response to the node threshold not meeting the requirement of the process principle, the node threshold is modified.

[0026] Optionally, the step of obtaining a target process node corresponding to the inclusion in the continuous casting slab based on the modified decision tree model includes:

[0027] Converting the modified decision tree model into a Bayesian network;

[0028] Wherein, the nodes of the Bayesian network correspond to the nodes of the decision tree model;

[0029] Obtaining a conditional probability table corresponding to all nodes in the Bayesian network;

[0030] Wherein, the conditional probability table includes the conditional probability corresponding to each node in the Bayesian network;

[0031] Selecting a target path from the initial paths of the Bayesian network based on the conditional probability table;

[0032] The initial path is a path from the root node to the leaf node of the Bayesian network, and the inclusion state data of the leaf node is inclusion;

[0033] Based on the target path, the target process node corresponding to the inclusion of the continuous casting billet is obtained.

[0034] Optionally, the step of selecting a target path from the initial paths of the Bayesian network based on the conditional probability table includes:

[0035] Based on the conditional probability table, obtaining an initial probability corresponding to each initial path in the Bayesian network;

[0036] The target path is selected from the initial paths based on the initial probability.

[0037] Optionally, the step of selecting the target path from the initial paths based on the initial probability includes:

[0038] The initial path with the highest initial probability is selected as the target path.

[0039] Optionally, the step of obtaining a target process node corresponding to the inclusion of the continuous casting slab based on the target path includes:

[0040] Modifying the node determination condition corresponding to each node in the target path respectively;

[0041] The node determination condition is used to characterize the size relationship between the initial process node of the node and the corresponding node threshold;

[0042] Based on the modified node determination condition, respectively obtain the probability gap value corresponding to each node;

[0043] The probability gap value is the gap between the initial probability corresponding to the target path after the node determination condition is modified and the initial probability corresponding to the target path before the node determination condition is modified;

[0044] Selecting a target node from all nodes on the target path based on the probability gap value;

[0045] Based on the target node, the target process node corresponding to the inclusion of the continuous casting billet is obtained.

[0046] Optionally, the step of selecting a target node from all nodes of the target path based on the probability gap value includes:

[0047] Sorting all nodes of the target path based on the probability gap value to obtain a sorting result;

[0048] Based on the sorting result, the target node is selected.

[0049] The present disclosure further provides a system for determining a process node of inclusions in a continuous casting slab, the system comprising:

[0050] A sample data acquisition module is used to acquire sample data of several billets during the continuous casting process;

[0051] The sample data includes process data and corresponding inclusion status data, and the process data includes a number of initial process nodes and corresponding actual values of the nodes;

[0052] A model acquisition module, configured to obtain a decision tree model based on the sample data;

[0053] The decision tree model includes leaf nodes and non-leaf nodes, the leaf nodes include the inclusion state data, and the non-leaf nodes include the initial process nodes and corresponding node thresholds;

[0054] A model modification module, used to modify the decision tree model based on preset standards;

[0055] A process node acquisition module is used to obtain the target process node corresponding to the inclusion of the continuous casting billet based on the modified decision tree model.

[0056] Optionally, in response to the preset standard including historical experience data of the continuous casting process, the model modification module includes:

[0057] A first threshold modification unit, configured to modify the node threshold of the non-leaf node in the decision tree model based on the historical experience data;

[0058] and / or,

[0059] A process node modification unit, configured to modify the initial process nodes of the non-leaf nodes in the decision tree model based on the historical experience data;

[0060] A model modification unit, configured to modify the decision tree model based on the modified initial process node;

[0061] and / or,

[0062] a deleting unit, configured to delete part of the decision paths of the decision tree model and / or part of the subtrees of the decision tree model based on the historical experience data;

[0063] and / or,

[0064] A path adding unit, configured to add a decision path from a root node of the decision tree model based on the historical experience data;

[0065] and / or,

[0066] In response to the preset standard including the process principle of the continuous casting process, the model modification module includes:

[0067] a determination unit, configured to determine whether the node threshold in the decision path of the decision tree model meets the requirements of the process principle;

[0068] The second threshold value modifying unit is configured to modify the node threshold value in response to the node threshold value not meeting the requirement of the process principle.

[0069] Optionally, the process node acquisition module includes:

[0070] A model conversion unit, used to convert the modified decision tree model into a Bayesian network;

[0071] Wherein, the nodes of the Bayesian network correspond to the nodes of the decision tree model;

[0072] A probability table unit, used to obtain conditional probability tables corresponding to all nodes in the Bayesian network;

[0073] Wherein, the conditional probability table includes the conditional probability corresponding to each node in the Bayesian network;

[0074] a target path selection unit, configured to select a target path from the initial paths of the Bayesian network based on the conditional probability table;

[0075] The initial path is a path from the root node to the leaf node of the Bayesian network, and the inclusion state data of the leaf node is inclusion;

[0076] A process node acquisition unit is used to obtain the target process node corresponding to the inclusion of the continuous casting billet based on the target path.

[0077] Optionally, the target path selection unit includes:

[0078] An initial probability acquisition subunit, configured to obtain an initial probability corresponding to each initial path in the Bayesian network based on the conditional probability table;

[0079] The target path selection subunit is configured to select the target path from the initial paths based on the initial probability.

[0080] Optionally, the target path selection subunit is further configured to select the initial path with the highest initial probability as the target path.

[0081] Optionally, the process node acquisition unit includes:

[0082] A determination condition modification subunit, configured to modify the node determination condition corresponding to each node in the target path;

[0083] The node determination condition is used to characterize the size relationship between the initial process node of the node and the corresponding node threshold;

[0084] A gap value obtaining subunit is used to obtain the probability gap value corresponding to each node based on the modified node determination condition;

[0085] The probability gap value is the gap between the initial probability corresponding to the target path after the node determination condition is modified and the initial probability corresponding to the target path before the node determination condition is modified;

[0086] a target node selection subunit, configured to select a target node from all nodes in the target path based on the probability gap value;

[0087] The process node acquisition subunit is used to obtain the target process node corresponding to the inclusion of the continuous casting billet based on the target node.

[0088] Optionally, the target node selection subunit is further configured to sort all nodes of the target path based on the probability gap value to obtain a sorting result;

[0089] Based on the sorting result, the target node is selected.

[0090] The present disclosure also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and used to run on the processor, wherein the processor implements the above-mentioned method for determining the process node of inclusions in the continuous casting billet when executing the computer program.

[0091] The present disclosure also provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the method for determining the process node of inclusions in the continuous casting billet described above is implemented.

[0092] The present disclosure also provides a computer program product, including a computer program, which, when executed by a processor, implements the method for determining a process node of inclusions in a continuous casting slab as described above.

[0093] On the basis of conforming to the common sense in this field, the above-mentioned preferred conditions can be arbitrarily combined to obtain the preferred embodiments of the present disclosure.

[0094] The positive progress of this disclosure is:

[0095] The present disclosure obtains a decision tree model through sample data of several billets in the continuous casting process, and then modifies the decision tree model according to preset standards to obtain the target process node corresponding to the inclusion of the continuous casting billet according to the modified decision tree model. It conveniently combines preset standards such as expert judgment experience with the root cause judgment rules mined from data, realizes the location of the root cause of inclusion, and accurately identifies the root cause of the inclusion problem. It is suitable for scenarios with large data volume, complex problems and the need to be combined with the background of the steel production industry. It has high credibility and is convenient for process personnel to improve the process according to the target process node to reduce the probability of inclusion. BRIEF DESCRIPTION OF THE DRAWINGS

[0096] Figure 1 This is a flow chart of a method for determining a process node of inclusions in a continuous casting slab according to Example 1 of the present disclosure;

[0097] Figure 2 This is a flow chart of step S14 in the method for determining a process node of inclusions in a continuous casting slab according to Example 1 of the present disclosure;

[0098] Figure 3 This is a flowchart of step S143 in the method for determining a process node of inclusions in a continuous casting slab according to Example 1 of the present disclosure;

[0099] Figure 4 This is a flow chart of step S144 in the method for determining a process node of inclusions in a continuous casting slab according to Example 1 of the present disclosure;

[0100] Figure 5 This is a flowchart of step S1443 in the method for determining the process node of inclusions in the continuous casting slab according to Example 1 of the present disclosure;

[0101] Figure 6 This is a specific example diagram of a decision tree model in the method for determining a process node of inclusions in a continuous casting slab according to Example 1 of the present disclosure;

[0102] Figure 7 This is a specific example diagram of the Bayesian network in the method for determining the process node of inclusions in the continuous casting slab according to Example 1 of the present disclosure;

[0103] Figure 8 This is a schematic diagram of the first module of the system for determining the process node of inclusions in the continuous casting slab according to Example 2 of the present disclosure;

[0104] Figure 9 This is a schematic diagram of a second module of a system for determining a process node of inclusions in a continuous casting slab according to Example 2 of the present disclosure;

[0105] Figure 10This is a structural diagram of an electronic device according to embodiment 3 of the present disclosure. DETAILED DESCRIPTION

[0106] The present disclosure is further illustrated below by way of examples, but the present disclosure is not limited to the scope of the examples.

[0107] In the embodiments of the present disclosure, prefixes such as "first" and "second" are used only to distinguish different description objects and have no limiting effect on the position, order, priority, quantity or content of the described objects. In the embodiments of the present disclosure, the use of prefixes such as ordinal numbers to distinguish description objects does not constitute a restriction on the described objects. For the statement of the described objects, please refer to the description in the context of the embodiments, and the use of such prefixes should not constitute an unnecessary restriction. In addition, in the description of this embodiment, unless otherwise specified, the meaning of "plurality" is two or more.

[0108] Example 1

[0109] This embodiment provides a method for determining the process nodes of inclusions in continuous casting billets, such as Figure 1 As shown, the determination method includes:

[0110] S11, obtaining sample data of a plurality of billets during the continuous casting process;

[0111] The sample data includes process data and corresponding inclusion status data, and the process data includes several initial process nodes and corresponding actual values of the nodes;

[0112] S12. Obtain a decision tree model based on the sample data;

[0113] The decision tree model includes leaf nodes and non-leaf nodes. Leaf nodes include inclusion status data, and non-leaf nodes include initial process nodes and corresponding node thresholds.

[0114] S13, modifying the decision tree model based on preset criteria;

[0115] S14. Based on the modified decision tree model, the target process node corresponding to the inclusion of the continuous casting billet is obtained.

[0116] Specifically, the blank is, for example, a slab. The process data of each slab during the production process and the label of whether there are inclusions, that is, the sample data, are collected. The label of whether there are inclusions, that is, the inclusion status data, includes two states: inclusion and normal. The process data includes: process data that may cause inclusions, such as: immersion nozzle depth, crystallizer water flow, superheat, etc.; slab production parameters determined according to production requirements, such as: aluminum content, silicon content, slab thickness, etc.; sensor collection data, such as: crystallizer liquid level fluctuation, temperature, etc. The collected process data are aligned according to the slab number as input data, that is, the process data of each slab. The output data corresponding to the input data is the label of whether there are inclusions.

[0117] After obtaining sample data of a number of billets in the continuous casting process, the sample data may be preprocessed; and a decision tree model may be obtained based on the preprocessed sample data.

[0118] The preprocessing includes at least one of data cleaning and data resampling.

[0119] Data cleaning includes: deleting abnormal data beyond three times the standard deviation; normalizing input data.

[0120] For each process data item, calculate the mean and standard deviation of the process data item. If the distance between the process data item of a sample data item and the mean exceeds three times the standard deviation, the sample data is considered to be abnormal data and deleted.

[0121] The input data is normalized using the minimum-maximum method. That is, for each process data item, the minimum and maximum values of the process data item are taken. For the observed value of the process data of each sample data, the value calculated by (observed value - minimum value) / (maximum value - minimum value) is used as the normalized observed value.

[0122] The mean, standard deviation, minimum, maximum, and observed value of each process data item here refer to the actual node value corresponding to the initial process node. For example, if the initial process node is the submerged nozzle depth, the corresponding actual node value is the actual value of the submerged nozzle depth.

[0123] Data resampling is performed based on the number of data with inclusions and the number of normal data without inclusions. Data with inclusions (i.e., data in the inclusion state) is considered as the impure sample data, while normal data without inclusions (i.e., data in the inclusion state) is considered as the normal sample data. The number of data is the number of corresponding sample data, which can be counted as the number of slabs.

[0124] The goal of resampling is to make the ratio between the number of mixed data and the number of normal data without mixed data not too different after sampling, for example, to control the ratio between the number of mixed data and the number of normal data without mixed data to about one to ten.

[0125] This can be achieved by upsampling and downsampling the mixed and clean normal data. For example, if the ratio between the mixed and clean normal data is set to 1:9, and the total number of sample data is 10,000, 1,000 mixed data items and 9,000 clean normal data items are required. If the number of mixed data items exceeds 1,000, downsampling is used, i.e., 1,000 mixed data items are randomly sampled without repetition. If the number of mixed data items is less than 1,000, upsampling is used, i.e., 1,000 mixed data items are randomly sampled with repetition. Similarly, if the number of clean normal data items exceeds 9,000, downsampling is used, i.e., 9,000 clean normal data items are randomly sampled without repetition. If the number of clean normal data items is less than 9,000, upsampling is used, i.e., 9,000 clean normal data items are randomly sampled with repetition. The values 10000, 1000, and 9000 are just examples and can be set or adjusted according to actual conditions.

[0126] Based on the preprocessed sample data, a decision tree model is obtained.

[0127] A tree-structured decision model, also known as a decision tree model, is used. A decision tree model is a binary classification model that predicts labels. It is interpretable and can be used to analyze the rationality of the method for determining process nodes for inclusions in continuous casting slabs from a process perspective. Since the input variables include both discrete and continuous variables, a Categorical and Regression Tree (CART) is typically used to construct the decision tree. The input variables are process data, with discrete variables representing discrete values at the nodes representing process data and continuous variables representing continuous values at the nodes representing process data. However, CART generally does not limit the number of times a continuous variable can appear in each branch. This may result in the same continuous variable appearing multiple times in a branch, making the method for determining process nodes for inclusions in continuous casting slabs redundant. Therefore, a special construction method is used to construct the decision tree model, requiring that each input variable appear at most once in each decision path. A decision tree model consists of several decision paths, which are paths from the root node to the leaf nodes of the decision tree model. Different nodes in each decision path correspond to different initial process nodes. The nodes of the decision tree model include leaf nodes and non-leaf nodes, and non-leaf nodes include the root node.

[0128] The logic of constructing a decision tree model includes initialization, termination conditions, building a constructor, and recursively building subtrees.

[0129] Initialization is to read in a data set and divide it into input variables and output variables. The data set is the sample data, the input variables are the process data, and the output variables are the inclusion status data.

[0130] Termination conditions include:

[0131] If all samples belong to the same category, there is no further splitting and the current node becomes a leaf node;

[0132] If the number of samples of the current node is less than the minimum partition threshold, the majority category of the node is directly returned;

[0133] There are no available features for further splitting, so the majority class for the node is returned.

[0134] When building a decision tree model, a termination condition, or stopping condition, is used to determine when to stop the tree's growth to avoid overfitting or overcomplication. A category is the discrete value of the target variable in a classification problem. A decision tree clusters samples into the same category by splitting features. The target variable is the output variable to be predicted, i.e., the data in the inclusion state.

[0135] Constructing the constructor: For each node, consider the path from that node back to the root node. All variables that appeared on this path must not appear again. The set of available input variables is the total set of variables minus the variables that appeared on this path. For all available input variables, use the maximum gain or Gini coefficient to find the optimal variable and the corresponding optimal split point, which will be used as the child node.

[0136] Recursively build subtrees: call the constructor for each subtree until the termination condition is triggered.

[0137] Using this method, a decision tree is constructed that satisfies the requirement that each variable appear at most once on each decision path. This allows for comprehensive consideration of the impact of all variables on each decision path while also avoiding the problem of redundant decision paths caused by repeated use of the same variable.

[0138] After obtaining the decision tree model, it is modified based on preset standards. Due to factors such as data acquisition, data distribution, and sensor errors, decision tree models based on data-driven mining may not fully meet the needs of process engineers. Modifications to the decision tree model based on expert experience data and / or process principles are necessary. The decision tree model is highly editable, allowing for modifications. Preset standards include historical experience data and / or process principles of the continuous casting process. Historical experience data includes expert experience data. Using expert experience data and process principles to make minor modifications to the decision tree model ensures that the decision path of the decision tree model meets the requirements of process engineers, improving the accuracy and reliability of the decision tree model.

[0139] Based on the modified decision tree model, the target process nodes corresponding to the inclusions in the continuous casting billet are obtained.

[0140] In this solution, a decision tree model is obtained through sample data of several billets in the continuous casting process, and then the decision tree model is modified according to preset standards to obtain the target process node corresponding to the inclusion of the continuous casting billet according to the modified decision tree model. It is convenient to combine preset standards such as expert judgment experience with the root cause judgment rules mined by data, realize the root cause positioning of inclusions, and accurately identify the root cause of the inclusion problem. It is suitable for scenarios with large data volume, complex problems and the need to be combined with the background of the steel production industry. It has high credibility and is convenient for process personnel to improve the process according to the target process node to reduce the probability of inclusions.

[0141] In one practicable solution, in response to the preset criteria including historical experience data of the continuous casting process, step S13 includes:

[0142] Based on historical experience data, the node thresholds of non-leaf nodes in the decision tree model are modified.

[0143] For example, to change the node splitting condition "pulling speed <= 1.5 m / min" to "pulling speed <= 1.4 m / min", this can be achieved by modifying the node judgment condition of the corresponding node. That is, non-leaf nodes also include node judgment conditions. The node judgment condition is used to characterize the size relationship between the initial process node of the node and the corresponding node threshold.

[0144] In this solution, based on historical experience data, the node thresholds of non-leaf nodes in the decision tree model are modified, thereby improving the accuracy and reliability of the decision tree model.

[0145] In one practicable solution, in response to the preset criteria including historical experience data of the continuous casting process, step S13 includes:

[0146] Based on historical experience data, the initial process nodes of non-leaf nodes in the decision tree model are modified;

[0147] Based on the modified initial process node, the decision tree model is modified.

[0148] Specifically, the splitting feature "superheat" of a node is modified to "mold water flow rate", and then the split point and subtree of the branch are reconstructed based on the new splitting feature. In other words, the node feature is replaced.

[0149] In this solution, the initial process nodes in the decision tree model are modified based on historical experience data, and then the decision tree model is further modified to improve the accuracy and reliability of the decision tree model.

[0150] In one practicable solution, in response to the preset criteria including historical experience data of the continuous casting process, step S13 includes:

[0151] Based on historical experience data, some decision paths of the decision tree model and / or some subtrees of the decision tree model are deleted.

[0152] Specifically, redundant split paths are deleted and redundant subtrees are removed to force the merging of decision paths.

[0153] In this solution, based on historical experience data, some decision paths and some subtrees of the decision tree model are deleted, thereby improving the accuracy and reliability of the decision tree model.

[0154] In one practicable solution, in response to the preset criteria including historical experience data of the continuous casting process, step S13 includes:

[0155] Based on historical experience data, a decision path is added from the root node of the decision tree model.

[0156] Specifically, a decision path determined by expert rules is added to the root node of the decision tree.

[0157] In this solution, based on historical experience data, a decision path is added from the root node of the decision tree model, which improves the accuracy and reliability of the decision tree model.

[0158] In one feasible solution, in response to the preset standard including the process principle of the continuous casting process, step S13 includes:

[0159] Determine whether the node thresholds in the decision path of the decision tree model meet the requirements of the process principle;

[0160] In response to the node threshold not meeting the requirement of the process principle, the node threshold is modified.

[0161] Specifically, the process principle establishes boundaries for process parameters. For example, superheat must be between 35 and 50 degrees to ensure normal production. The decision tree model examines the discriminant values for the superheat process parameter in all decision paths and verifies their discriminant values. The discriminant value represents the state of the inclusion status data, either normal or impure. The discriminant value is the node threshold. If the discriminant value is normal and the required discriminant value is not between 35 and 50, the decision path must be modified and the node threshold adjusted.

[0162] In this solution, when the node threshold in the decision path of the decision tree model does not meet the requirements of the process principle, the node threshold is modified to improve the accuracy and reliability of the decision tree model.

[0163] In one feasible solution, Figure 2 As shown, step S14 includes:

[0164] S141, converting the modified decision tree model into a Bayesian network;

[0165] Among them, the nodes of the Bayesian network correspond to the nodes of the decision tree model;

[0166] S142, obtaining a conditional probability table corresponding to all nodes in the Bayesian network;

[0167] Among them, the conditional probability table includes the conditional probability corresponding to each node in the Bayesian network;

[0168] S143. Selecting a target path from the initial paths of the Bayesian network based on the conditional probability table;

[0169] Among them, the initial path is the path from the root node of the Bayesian network to the leaf node, and the inclusion state data of the leaf node is inclusion;

[0170] S144. Based on the target path, obtain the target process node corresponding to the inclusion of the continuous casting billet.

[0171] Specifically, first convert the modified decision tree into a Bayesian network structure. Since the tree structure is naturally a network structure, the conversion of the decision tree model can be achieved through the following methods:

[0172] Each feature and target variable in the decision tree model is used as a node of the Bayesian network; the features and target variables are the nodes of the decision tree model;

[0173] Convert the root node of the decision tree model into the root node of the Bayesian network;

[0174] Add edges to the Bayesian network based on the parent-child relationships in the decision tree model. Because a decision tree model is a binary classification tree, each internal node and leaf node has exactly one parent node. A Bayesian network is a directed graph. For example, if the features include a first feature M and a second feature N, and if the first feature M is the parent of the second feature N, then add an edge from the first feature M to the second feature N in the Bayesian network.

[0175] The above method is used to transform the network structure of the decision tree model into the corresponding Bayesian network structure.

[0176] A Bayesian network is determined by two parts: the network structure and the conditional probability table. Therefore, it is necessary to determine the conditional probability table for the Bayesian network. This can be determined using data-based maximum likelihood estimation, which calculates the conditional probability of each node in the Bayesian network based on sample data.

[0177] The conditional probability distribution of the root node of a Bayesian network is estimated using sample data. Assuming the root node is Z, the conditional probability of the root node is P(Z=z) = (Count(Z=z)) / Total number of samples. Here, z represents the node value corresponding to the initial process node of node Z. This node value can be the actual node value of the initial process node or the predicted node value of the initial process node. The predicted node value is set or adjusted based on actual conditions.

[0178] The conditional probability distributions of internal and leaf nodes in a Bayesian network are estimated using the splitting rule of the decision tree model and sample data. The conditional probability of each node in a Bayesian network depends only on its parent node, so only the parent nodes of internal and leaf nodes need to be considered. The nodes in a Bayesian network consist of a first node X and a second node Y. If the parent of first node X is second node Y, then the conditional probability of first node X, P(X=x|Y=y), equals (count(X=x, Y=y)) / (count(Y=y)). To avoid zero probabilities, a smoothing term is introduced: P(X=x|Y=y)=(count(X=x, Y=y)+c) / (count(Y=y)+k×c), where k is the number of possible values for first node X, which can be 2, and c is a fixed positive number. The value of 2 for k is just an example and can be set or adjusted based on actual conditions. x represents the node value corresponding to the initial process node of first node X, and y represents the node value corresponding to the initial process node of second node Y.

[0179] The conditional probability table of the Bayesian network is obtained through the above method.

[0180] The Bayesian network includes multiple initial paths. The target path is selected from the initial paths according to the conditional probability table, and then the target process node corresponding to the inclusion of the continuous casting billet is obtained.

[0181] In this scheme, the modified decision tree model is converted into a Bayesian network to obtain the conditional probability table corresponding to all nodes in the Bayesian network. Then, the target path is selected based on the conditional probability table to obtain the target process node corresponding to the inclusion of the continuous casting billet, thereby ensuring the accuracy and reliability of the target process node.

[0182] In one feasible solution, Figure 3 As shown, step S143 includes:

[0183] S1431. Based on the conditional probability table, obtain the initial probability corresponding to each initial path in the Bayesian network;

[0184] S1432. Based on the initial probability, select a target path from the initial paths.

[0185] Specifically, since it is necessary to identify the root cause of the inclusion problem in the billet, that is, to locate the root cause, the leaf node of the initial path is the inclusion status data of "whether inclusion occurs". All process data are known, and the inclusion status data corresponding to the process data include data with the status of inclusion. All initial paths that will cause inclusions can be found, and then the initial probability of each initial path can be calculated. For example, an initial path is A -> B -> C -> D -> E -> inclusion, where the process data A=a, B=b, C=c, D=d, and E=e. Among them, A, B, C, D, and E represent nodes in the Bayesian network, and a, b, c, d, and e represent the node values corresponding to the initial process nodes of A, B, C, D, and E, respectively.

[0186] According to the initial path, the initial probability P (A=a, B=b, C=c, D=d, E=e|inclusion) corresponding to the initial path can be obtained.

[0187] P(A=a, B=b, C=c, D=d, E=e|inclusion) ∝P(A=a)×P(B=b|A=a)×P(C=c|A=a, B=b)×P(D= d|A=a, B=b, C=c)×P (E=e|A=a, B=b, C=c, D=d)×P (inclusion|A=a, B=b, C=c, D=d, E=e).

[0188] Here, P(A=a), P(B=b|A=a), P(C=c|A=a, B=b), P(D=d|A=a, B=b, C=c), P(E=e|A=a, B=b, C=c, D=d), and P(Inclusion|A=a, B=b, C=c, D=d, E=e) represent the conditional probabilities of the corresponding nodes. The conditional probabilities of nodes in a Bayesian network can be obtained using the instructions for obtaining a conditional probability table, and are not repeated here.

[0189] The initial probability of the initial path can be calculated through the conditional probability table of the Bayesian network.

[0190] Through the graph structure of the Bayesian network, it is easy to search for all initial paths from the root node to the leaf node of "whether inclusion occurs", and then it is also easy to search for all initial paths where the inclusion status data of the leaf node is inclusion.

[0191] According to the initial probability, the target path is selected from all the initial paths.

[0192] In this scheme, the initial probability corresponding to each initial path in the Bayesian network is obtained through the conditional probability table, and then the target path is selected from the initial paths, ensuring the accuracy and reliability of the target path.

[0193] In one feasible solution, step S1432 includes:

[0194] The initial path with the highest initial probability is selected as the target path.

[0195] Specifically, the path with the highest initial probability among all the initial paths is the root cause path that causes inclusion, and the root cause path is used as the target path.

[0196] In this scheme, the accuracy and reliability of the target path are guaranteed by selecting the initial path with the highest initial probability as the target path.

[0197] In one feasible solution, Figure 4 As shown, step S144 includes:

[0198] S1441, modify the node determination condition corresponding to each node in the target path respectively;

[0199] Among them, the node determination condition is used to characterize the size relationship between the initial process node of the node and the corresponding node threshold;

[0200] S1442. Based on the modified node determination condition, obtain the probability gap value corresponding to each node;

[0201] The probability gap value is the gap between the initial probability of the target path after the node determination condition is modified and the initial probability of the target path before the node determination condition is modified.

[0202] S1443. Select a target node from all nodes on the target path based on the probability gap value;

[0203] S1444. Based on the target node, obtain the target process node corresponding to the continuous casting slab inclusion.

[0204] Specifically, if the root cause path, i.e., the target path, is A -> B -> C -> D -> E -> inclusion, then the following steps are used to retrieve the root cause node, which is the target process node:

[0205] Traverse A, B, C, D, and E, and for each node, change the node decision condition. Each node's node decision condition is either yes or no, represented by a value of 0 or 1. If the node decision condition is 0, change it to 1; if the node decision condition is 1, change it to 0. After flipping the node decision condition, recalculate the initial probability of the target path. Probability gap value = (absolute value (initial probability of the target path after modifying the node decision condition - initial probability of the target path before modifying the node decision condition)) / initial probability of the target path before modifying the node decision condition. Record the probability gap value for each node.

[0206] After the traversal is completed, the target node is selected from all nodes of the target path based on the probability gap value, and the initial process node of the target node is used as the target process node corresponding to the continuous casting billet inclusion.

[0207] In this scheme, the node judgment conditions corresponding to each node in the target path are modified to obtain the probability gap value corresponding to each node respectively, and then the target node is selected from all the nodes of the target path according to the probability gap value to obtain the target process node corresponding to the continuous casting billet inclusion, thereby ensuring the accuracy and reliability of the target node and target process node acquisition.

[0208] In one feasible solution, Figure 5 As shown, step S1443 includes:

[0209] S14431. Sort all nodes on the target path based on the probability gap value to obtain a sorting result;

[0210] S14432. Select a target node based on the sorting result.

[0211] Specifically, the probability gap values are sorted from largest to smallest to obtain a probability gap ranking result, which is equivalent to sorting the nodes on the target path corresponding to the probability gap values. The top K nodes in the ranking result are selected as the root cause nodes. K is a positive integer, and the size of K can be determined by the process personnel.

[0212] In this scheme, all nodes of the target path are sorted based on the probability gap value to obtain the sorting result, and the target node is selected based on the sorting result, which ensures the accuracy and reliability of the target node.

[0213] The working principle of the method for determining the process node of inclusions in continuous casting slabs of this embodiment is described below with reference to specific examples:

[0214] The sample data of the continuous casting process in the steel plant was analyzed to obtain a decision tree model for determining inclusions in the continuous casting billet. Figure 6 As shown in the figure, the root node of the decision tree model is "Casting Speed < 95," where "Casting Speed" is the initial process node and "95" is the node threshold. The leaf nodes of the decision tree model are "Normal" and "Inclusions." The non-leaf nodes of the decision tree model include the root node and internal nodes. Internal nodes include "Casting Speed Fluctuation <= 0.04," "RH-oxygen Blowing (RH-oxygen blowwing, a method of removing carbon from molten steel) Total Oxygen Content <= 25," "Submerged Nozzle Depth <= 141," "Superheat <= 38," and "Aluminum Content <= 300." The left branch of the decision tree model indicates that the judgment condition is met, while the right branch indicates that the judgment condition is not met.

[0215] This decision tree model uses a special structure. Each process data point appears at most once and only once in the decision path of the decision tree, simplifying the structure of the decision tree. An overly complex decision tree structure will lead to an overly complex structure in the subsequent Bayesian network, affecting the solution of conditional probabilities.

[0216] The decision tree model is modified based on preset criteria such as expert judgment experience.

[0217] Convert the modified decision tree model into a Bayesian network. Figure 7 As shown in Figure 1, the root node of the Bayesian network is “casting speed”. The nodes of the Bayesian network also include “casting speed fluctuation”, “RHOB total oxygen content”, “immersion nozzle depth”, “superheat”, “aluminum content”, and “inclusions”.

[0218] The conditional probability of each node in the Bayesian network is calculated through the distribution of sample data to obtain a conditional probability table; based on the conditional probability table, the initial probability corresponding to each initial path in the Bayesian network is calculated, and the initial path with the largest initial probability is selected as the target path.

[0219] On the root cause path, the node judgment conditions of each node are flipped one by one to find the target node that has the greatest impact on inclusions. The initial process node of the target node is the root cause node.

[0220] In this embodiment, a decision tree model is obtained through sample data of several billets in the continuous casting process, and then the decision tree model is modified according to preset standards to obtain the target process node corresponding to the inclusion of the continuous casting billet according to the modified decision tree model. The preset standards such as expert judgment experience are conveniently combined with the root cause judgment rules mined from data, thereby realizing the location of the root cause of inclusions and accurately identifying the root cause of the inclusion problem. It is suitable for scenarios with large data volumes, complex problems and the need to be combined with the background of the steel production industry. It has high credibility and is convenient for process personnel to improve the process according to the target process node to reduce the probability of inclusions.

[0221] Example 2

[0222] Corresponding to the aforementioned embodiment of the method for determining a process node of inclusions in a continuous casting slab, the present disclosure also provides an embodiment of a system for determining a process node of inclusions in a continuous casting slab.

[0223] like Figure 8 As shown, the determination system includes:

[0224] The sample data acquisition module 1 is used to acquire sample data of a plurality of billets during the continuous casting process;

[0225] The sample data includes process data and corresponding inclusion status data, and the process data includes several initial process nodes and corresponding actual values of the nodes;

[0226] Model acquisition module 2, used to obtain a decision tree model based on sample data;

[0227] The decision tree model includes leaf nodes and non-leaf nodes. Leaf nodes include inclusion status data, and non-leaf nodes include initial process nodes and corresponding node thresholds.

[0228] Model modification module 3, used to modify the decision tree model based on preset standards;

[0229] The process node acquisition module 4 is used to obtain the target process node corresponding to the inclusion of the continuous casting billet based on the modified decision tree model.

[0230] In one embodiment, the preset criteria include historical experience data of the continuous casting process, such as Figure 9 As shown, the model modification module 3 includes:

[0231] A first threshold modification unit 31 is used to modify the node threshold of non-leaf nodes in the decision tree model based on historical experience data;

[0232] and / or,

[0233] A process node modification unit 32 is used to modify the initial process nodes of non-leaf nodes in the decision tree model based on historical experience data;

[0234] A model modification unit 33 is used to modify the decision tree model based on the modified initial process node;

[0235] and / or,

[0236] A deleting unit 34 is configured to delete part of the decision paths of the decision tree model and / or part of the subtrees of the decision tree model based on historical experience data;

[0237] and / or,

[0238] A path adding unit 35 is used to add a decision path from the root node of the decision tree model based on historical experience data;

[0239] and / or,

[0240] In response to the preset standards including the process principle of the continuous casting process, the model modification module 3 includes:

[0241] A determination unit 36 is used to determine whether the node threshold in the decision path of the decision tree model meets the requirements of the process principle;

[0242] The second threshold modifying unit 37 is configured to modify the node threshold in response to the node threshold not meeting the requirement of the process principle.

[0243] In one feasible solution, the process node acquisition module 4 includes:

[0244] A model conversion unit 41 is used to convert the modified decision tree model into a Bayesian network;

[0245] Among them, the nodes of the Bayesian network correspond to the nodes of the decision tree model;

[0246] The probability table unit 42 is used to obtain the conditional probability table corresponding to all nodes in the Bayesian network;

[0247] Among them, the conditional probability table includes the conditional probability corresponding to each node in the Bayesian network;

[0248] a target path selection unit 43, configured to select a target path from the initial paths of the Bayesian network based on the conditional probability table;

[0249] Among them, the initial path is the path from the root node of the Bayesian network to the leaf node, and the inclusion state data of the leaf node is inclusion;

[0250] The process node acquisition unit 44 is used to obtain the target process node corresponding to the inclusion of the continuous casting billet based on the target path.

[0251] In one feasible solution, the target path selection unit 43 includes:

[0252] The initial probability acquisition subunit 431 is used to obtain the initial probability corresponding to each initial path in the Bayesian network based on the conditional probability table;

[0253] The target path selection subunit 432 is configured to select a target path from the initial paths based on the initial probability.

[0254] In an implementable solution, the target path selection subunit 432 is further configured to select an initial path with the highest initial probability as the target path.

[0255] In one feasible solution, the process node acquisition unit 44 includes:

[0256] The determination condition modification subunit 441 is used to modify the node determination condition corresponding to each node in the target path respectively;

[0257] Among them, the node determination condition is used to characterize the size relationship between the initial process node of the node and the corresponding node threshold;

[0258] The gap value obtaining subunit 442 is used to obtain the probability gap value corresponding to each node based on the modified node determination condition;

[0259] The probability gap value is the gap between the initial probability of the target path after the node determination condition is modified and the initial probability of the target path before the node determination condition is modified.

[0260] a target node selection subunit 443 for selecting a target node from all nodes in the target path based on the probability gap value;

[0261] The process node acquisition subunit 444 is used to obtain the target process node corresponding to the continuous casting slab inclusion based on the target node.

[0262] In one feasible solution, the target node selection subunit 443 is further configured to sort all nodes of the target path based on the probability gap value to obtain a sorting result;

[0263] Based on the sorting results, the target node is selected.

[0264] In this embodiment, a decision tree model is obtained through sample data of several billets in the continuous casting process, and then the decision tree model is modified according to preset standards to obtain the target process node corresponding to the inclusion of the continuous casting billet according to the modified decision tree model. The preset standards such as expert judgment experience are conveniently combined with the root cause judgment rules mined from data, thereby realizing the location of the root cause of inclusions and accurately identifying the root cause of the inclusion problem. It is suitable for scenarios with large data volumes, complex problems and the need to be combined with the background of the steel production industry. It has high credibility and is convenient for process personnel to improve the process according to the target process node to reduce the probability of inclusions.

[0265] Since the system embodiments generally correspond to the method embodiments, reference will be made to the description of the method embodiments for relevant details. The system embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separate, and the components of the units may or may not be physical units, i.e., they may be located in one place or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of the disclosed solution.

[0266] Example 3

[0267] Figure 10 This is a structural schematic diagram of an electronic device showing an example embodiment of the present disclosure, the electronic device includes a memory, a processor, and a computer program stored in the memory and for running on the processor, and when the processor executes the computer program, it implements the method for determining the process node of the continuous casting inclusion as described in any of the above embodiments. Figure 10 The electronic device 90 shown is only an example and should not limit the functionality and scope of use of the embodiments of the present disclosure.

[0268] like Figure 10 As shown, the electronic device 90 may be a general-purpose computing device, such as a server device. Components of the electronic device 90 may include, but are not limited to, the at least one processor 91, the at least one memory 92, and a bus 93 connecting different system components (including the memory 92 and the processor 91).

[0269] The bus 93 includes a data bus, an address bus, and a control bus.

[0270] The memory 92 may include a volatile memory, such as a random access memory (RAM) 921 and / or a cache memory 922 , and may further include a read-only memory (ROM) 923 .

[0271] The memory 92 may also include a program tool 925 (or utility) having a set (at least one) of program modules 924, such program modules 924 including but not limited to: an operating system, one or more application programs, other program modules and program data, each of which or some combination may include an implementation of a network environment.

[0272] The processor 91 executes various functional applications and data processing by running the computer program stored in the memory 92, such as the method for determining the process node of the continuous casting inclusion provided by any of the above embodiments.

[0273] The electronic device 90 can also communicate with one or more external devices 94 (e.g., a keyboard, pointing device, etc.). This communication can occur via an input / output (I / O) interface 95. Furthermore, the electronic device 90 can communicate with one or more networks (e.g., a local area network (LAN), a wide area network (WAN), and / or a public network, such as the Internet) via a network adapter 96. As shown, the network adapter 96 communicates with other modules of the electronic device 90 via a bus 93. It should be understood that, although not shown in the figure, other hardware and / or software modules can be used in conjunction with the electronic device 90, including but not limited to microcode, device drivers, redundant processors, external disk drive arrays, RAID (RAID) systems, tape drives, and data backup storage systems.

[0274] It should be noted that although several units / modules or sub-units / modules of the electronic device are mentioned in the detailed description above, this division is merely exemplary and not mandatory. In fact, according to the embodiments of the present disclosure, the features and functions of two or more units / modules described above can be embodied in one unit / module. Conversely, the features and functions of one unit / module described above can be further divided and embodied by multiple units / modules.

[0275] Example 4

[0276] The embodiments of the present disclosure further provide a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method for determining the process node of inclusions in a continuous casting slab provided in any of the above embodiments.

[0277] The readable storage medium may include, but is not limited to, a portable disk, a hard disk, a random access memory, a read-only memory, an erasable programmable read-only memory, an optical storage device, a magnetic storage device, or any suitable combination thereof.

[0278] Example 5

[0279] An embodiment of the present disclosure further provides a computer program product, comprising a computer program, which, when executed by a processor, implements any of the above-mentioned methods for determining a process node of inclusions in a continuous casting billet.

[0280] The program code for executing the computer program product of the present disclosure may be written in any combination of one or more programming languages, and the program code may be executed entirely on the user device, partially on the user device, as a standalone software package, partially on the user device and partially on a remote device, or entirely on the remote device.

[0281] While specific embodiments of the present disclosure have been described above, those skilled in the art will appreciate that these are merely illustrative and that the scope of protection of the present disclosure is defined by the appended claims. Those skilled in the art may make various changes or modifications to these embodiments without departing from the principles and essence of the present disclosure, and such changes and modifications are intended to fall within the scope of protection of the present disclosure.

Claims

1. A method for determining process nodes of inclusions in continuous casting billets, characterized in that: The determination method includes: Obtain sample data of several billets during the continuous casting process; The sample data includes process data and corresponding inclusion status data, and the process data includes a number of initial process nodes and corresponding actual values of the nodes; Based on the sample data, a decision tree model is obtained; The decision tree model includes leaf nodes and non-leaf nodes, the leaf nodes include the inclusion state data, and the non-leaf nodes include the initial process nodes and corresponding node thresholds; Modifying the decision tree model based on preset criteria; Based on the modified decision tree model, a target process node corresponding to the continuous casting slab inclusion is obtained; The step of obtaining the target process node corresponding to the inclusion in the continuous casting billet based on the modified decision tree model includes: Converting the modified decision tree model into a Bayesian network; Wherein, the nodes of the Bayesian network correspond to the nodes of the decision tree model; Obtaining a conditional probability table corresponding to all nodes in the Bayesian network; Wherein, the conditional probability table includes the conditional probability corresponding to each node in the Bayesian network; Selecting a target path from the initial paths of the Bayesian network based on the conditional probability table; The initial path is a path from the root node to the leaf node of the Bayesian network, and the inclusion state data of the leaf node is inclusion; Based on the target path, obtaining the target process node corresponding to the continuous casting slab inclusion; The step of obtaining a target process node corresponding to the inclusion of the continuous casting slab based on the target path includes: Modifying the node determination condition corresponding to each node in the target path respectively; The node determination condition is used to characterize the size relationship between the initial process node of the node and the corresponding node threshold; Based on the modified node determination condition, respectively obtain the probability gap value corresponding to each node; The probability gap value is the gap between the initial probability corresponding to the target path after the node determination condition is modified and the initial probability corresponding to the target path before the node determination condition is modified; Selecting a target node from all nodes on the target path based on the probability gap value; Based on the target node, the target process node corresponding to the inclusion of the continuous casting billet is obtained.

2. The method for determining the process node of inclusions in continuous casting according to claim 1, characterized in that: In response to the preset criteria including historical experience data of the continuous casting process, the step of modifying the decision tree model based on the preset criteria includes: Based on the historical experience data, modify the node threshold of the non-leaf node in the decision tree model; and / or, Based on the historical experience data, modifying the initial process nodes of the non-leaf nodes in the decision tree model; and modifying the decision tree model based on the modified initial process nodes; and / or, Based on the historical experience data, deleting some decision paths of the decision tree model and / or some subtrees of the decision tree model; and / or, Adding a decision path from the root node of the decision tree model based on the historical experience data; and / or, In response to the preset criteria including the process principle of the continuous casting process, the step of modifying the decision tree model based on the preset criteria includes: Determining whether the node threshold in the decision path of the decision tree model meets the requirements of the process principle; In response to the node threshold not meeting the requirement of the process principle, the node threshold is modified.

3. The method for determining the process node of inclusions in a continuous casting slab according to claim 1, characterized in that: The step of selecting a target path from the initial paths of the Bayesian network based on the conditional probability table includes: Based on the conditional probability table, obtaining an initial probability corresponding to each initial path in the Bayesian network; The target path is selected from the initial paths based on the initial probability.

4. The method for determining the process node of inclusions in a continuous casting slab according to claim 3, characterized in that: The step of selecting the target path from the initial paths based on the initial probability includes: The initial path with the highest initial probability is selected as the target path.

5. The method for determining the process node of inclusions in a continuous casting slab according to any one of claims 1, 3 and 4, characterized in that: The step of selecting a target node from all nodes of the target path based on the probability gap value includes: Sorting all nodes of the target path based on the probability gap value to obtain a sorting result; Based on the sorting result, the target node is selected.

6. A system for determining process nodes of inclusions in continuous casting billets, characterized in that: The determination system comprises: A sample data acquisition module is used to acquire sample data of several billets during the continuous casting process; The sample data includes process data and corresponding inclusion status data, and the process data includes a number of initial process nodes and corresponding actual values of the nodes; A model acquisition module, configured to obtain a decision tree model based on the sample data; The decision tree model includes leaf nodes and non-leaf nodes, the leaf nodes include the inclusion state data, and the non-leaf nodes include the initial process nodes and corresponding node thresholds; A model modification module, used to modify the decision tree model based on preset standards; A process node acquisition module, configured to obtain a target process node corresponding to the inclusion in the continuous casting billet based on the modified decision tree model; The process node acquisition module includes: A model conversion unit, used to convert the modified decision tree model into a Bayesian network; Wherein, the nodes of the Bayesian network correspond to the nodes of the decision tree model; A probability table unit, used to obtain conditional probability tables corresponding to all nodes in the Bayesian network; Wherein, the conditional probability table includes the conditional probability corresponding to each node in the Bayesian network; a target path selection unit, configured to select a target path from the initial paths of the Bayesian network based on the conditional probability table; The initial path is a path from the root node to the leaf node of the Bayesian network, and the inclusion state data of the leaf node is inclusion; A process node acquisition unit, configured to obtain the target process node corresponding to the inclusion of the continuous casting billet based on the target path; The process node acquisition unit includes: A determination condition modification subunit, configured to modify the node determination condition corresponding to each node in the target path; The node determination condition is used to characterize the size relationship between the initial process node of the node and the corresponding node threshold; A gap value obtaining subunit is used to obtain the probability gap value corresponding to each node based on the modified node determination condition; The probability gap value is the gap between the initial probability corresponding to the target path after the node determination condition is modified and the initial probability corresponding to the target path before the node determination condition is modified; a target node selection subunit, configured to select a target node from all nodes in the target path based on the probability gap value; The process node acquisition subunit is used to obtain the target process node corresponding to the inclusion of the continuous casting billet based on the target node.

7. An electronic device comprising a memory, a processor, and a computer program stored in the memory and configured to run on the processor, wherein: When the processor executes the computer program, the method for determining the process node of inclusions in the continuous casting slab according to any one of claims 1 to 5 is implemented.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for determining a process node of inclusions in a continuous casting slab according to any one of claims 1 to 5 is implemented.

9. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the method for determining a process node of inclusions in a continuous casting slab according to any one of claims 1 to 5 is implemented.

Images