CN117593254A - Refined ladle slag prediction method and system adopting regression decision tree - Google Patents
Refined ladle slag prediction method and system adopting regression decision tree Download PDFInfo
- Publication number
- CN117593254A CN117593254A CN202311438276.3A CN202311438276A CN117593254A CN 117593254 A CN117593254 A CN 117593254A CN 202311438276 A CN202311438276 A CN 202311438276A CN 117593254 A CN117593254 A CN 117593254A
- Authority
- CN
- China
- Prior art keywords
- data
- refined
- slag
- decision tree
- ladle
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 239000002893 slag Substances 0.000 title claims abstract description 128
- 238000003066 decision tree Methods 0.000 title claims abstract description 86
- 238000000034 method Methods 0.000 title claims abstract description 56
- 238000007670 refining Methods 0.000 claims abstract description 68
- 238000012549 training Methods 0.000 claims abstract description 48
- 238000010801 machine learning Methods 0.000 claims abstract description 28
- 230000008569 process Effects 0.000 claims abstract description 28
- 238000007664 blowing Methods 0.000 claims abstract description 24
- 229910000831 Steel Inorganic materials 0.000 claims description 25
- 239000010959 steel Substances 0.000 claims description 25
- 238000012545 processing Methods 0.000 claims description 23
- 238000007781 pre-processing Methods 0.000 claims description 21
- 238000010606 normalization Methods 0.000 claims description 18
- 230000002159 abnormal effect Effects 0.000 claims description 4
- 238000004140 cleaning Methods 0.000 claims description 3
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 abstract description 6
- 230000008878 coupling Effects 0.000 abstract description 2
- 238000010168 coupling process Methods 0.000 abstract description 2
- 238000005859 coupling reaction Methods 0.000 abstract description 2
- 230000003993 interaction Effects 0.000 abstract description 2
- 238000009847 ladle furnace Methods 0.000 abstract description 2
- 239000002245 particle Substances 0.000 description 11
- 238000010586 diagram Methods 0.000 description 10
- 238000012360 testing method Methods 0.000 description 9
- 238000004519 manufacturing process Methods 0.000 description 8
- 238000004590 computer program Methods 0.000 description 7
- 238000009826 distribution Methods 0.000 description 7
- 238000002474 experimental method Methods 0.000 description 7
- 230000006870 function Effects 0.000 description 7
- 229910052760 oxygen Inorganic materials 0.000 description 7
- QVGXLLKOCUKJST-UHFFFAOYSA-N atomic oxygen Chemical compound [O] QVGXLLKOCUKJST-UHFFFAOYSA-N 0.000 description 6
- 230000000694 effects Effects 0.000 description 6
- 238000009776 industrial production Methods 0.000 description 6
- 239000001301 oxygen Substances 0.000 description 6
- IJGRMHOSHXDMSA-UHFFFAOYSA-N Atomic nitrogen Chemical compound N#N IJGRMHOSHXDMSA-UHFFFAOYSA-N 0.000 description 4
- 238000011156 evaluation Methods 0.000 description 4
- 238000005192 partition Methods 0.000 description 4
- 238000005299 abrasion Methods 0.000 description 3
- 238000004422 calculation algorithm Methods 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 3
- 239000007789 gas Substances 0.000 description 3
- 239000012535 impurity Substances 0.000 description 3
- 229910052757 nitrogen Inorganic materials 0.000 description 3
- 238000004806 packaging method and process Methods 0.000 description 3
- 238000005096 rolling process Methods 0.000 description 3
- 238000003860 storage Methods 0.000 description 3
- XKRFYHLGVUSROY-UHFFFAOYSA-N Argon Chemical compound [Ar] XKRFYHLGVUSROY-UHFFFAOYSA-N 0.000 description 2
- 230000002411 adverse Effects 0.000 description 2
- 230000006399 behavior Effects 0.000 description 2
- 238000006243 chemical reaction Methods 0.000 description 2
- 239000002184 metal Substances 0.000 description 2
- 239000002994 raw material Substances 0.000 description 2
- 239000002699 waste material Substances 0.000 description 2
- 238000003723 Smelting Methods 0.000 description 1
- 230000004913 activation Effects 0.000 description 1
- 229910052786 argon Inorganic materials 0.000 description 1
- 238000004945 emulsification Methods 0.000 description 1
- 238000009851 ferrous metallurgy Methods 0.000 description 1
- 238000007667 floating Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000008140 language development Effects 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000012067 mathematical method Methods 0.000 description 1
- 238000013178 mathematical model Methods 0.000 description 1
- 238000005272 metallurgy Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 210000002569 neuron Anatomy 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000004321 preservation Methods 0.000 description 1
- 238000000746 purification Methods 0.000 description 1
- 230000001105 regulatory effect Effects 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
- 230000005514 two-phase flow Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/72—Data preparation, e.g. statistical preprocessing of image or video features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/766—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using regression, e.g. by projecting features on hyperplanes
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/52—Surveillance or monitoring of activities, e.g. for recognising suspicious objects
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30108—Industrial image inspection
- G06T2207/30136—Metal
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30232—Surveillance
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Artificial Intelligence (AREA)
- Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Databases & Information Systems (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Quality & Reliability (AREA)
- Treatment Of Steel In Its Molten State (AREA)
Abstract
The invention discloses a prediction method and a prediction system for refined ladle slag by adopting a regression decision tree. Refining data in the ladle bottom blowing refining process is collected, then the refining data is preprocessed, and the preprocessed data is input into a machine learning regression decision tree model for training and prediction, so that a prediction result is obtained. According to the invention, the refined ladle structure is converted into a corresponding machine learning special structure, unnecessary interaction and coupling are avoided, network parameters with similar characteristics are reused, the information of ladle refining physical principles is introduced, the dependence of a model on a sample is greatly reduced, and the modeling, training and learning efficiency and the model application success rate of a decision tree model can be remarkably improved. The decision tree training model is built for training water model experimental data and industrial field experimental data, so that the content of components in the ladle refining process and the condition in the ladle furnace under different conditions are judged, and the ladle in the working process can be predicted and monitored in real time.
Description
Technical Field
The invention relates to the technical field of ferrous metallurgy ladle refining, in particular to a refined ladle slag prediction method and a refined ladle slag prediction system adopting a regression decision tree.
Background
In the current molten steel smelting process, the molten steel smelted by a converter and an electric furnace still contains more O, N gas impurities and Ca, mg, si, al impurities, and bottom blowing ladle refining is widely applied as a refining mode with the most economical, effective and safe external refining.
The ladle slag prediction system is a complex part in the whole steel production system, and relates to complex physical and chemical changes in the ladle, the ladle is in a high-temperature environment in the operation process, the heat transfer of the ladle can cause other phase changes, and the burning high-temperature two-phase flow lacks effective testing means, so that the ladle slag prediction system is difficult to describe by a classical mathematical method, and no classical mature and reliable mathematical model exists at present.
The argon in the industrial site is blown into the molten steel through the air holes, and is mixed with the molten steel to be adsorbed to the slag layer in a floating way under the driving of the gas plume, so that the purpose of purifying the molten steel is achieved. The gas blows to the slag layer to form bulges to push slag to the surrounding area of the refining ladle to form slag eyes, and the slag eyes area exposes molten steel to the air to easily absorb nitrogen and oxygen and react with the nitrogen and oxygen, so that the molten steel is oxidized, the slag layer is blown away to be unfavorable for heat preservation, and the temperature of the molten steel is reduced to be unfavorable for removing impurities. In addition, slag-metal interface reactions at the periphery of the eyezone break down slag droplets, commonly referred to as "slag emulsification", which can increase the interface area of slag to metal reactions, can form slag inclusions, and can cause adverse effects such as slag entrainment. Therefore, timely prediction of the behaviors such as the size, the shape, the position and the like of the slag hole is particularly important to improving the refining effect.
Disclosure of Invention
The invention provides the prediction method and the prediction system for the refined ladle slag by adopting the regression decision tree, which can predict and monitor the ladle in real time in the working process, thereby effectively guiding industrial production.
The invention provides a prediction method for refined ladle slag by adopting a regression decision tree, which comprises the following steps:
collecting refining data in the ladle bottom blowing refining process;
preprocessing the refined data;
and inputting the preprocessed data into a machine learning regression decision tree model for training and predicting to obtain a prediction result.
Specifically, the collecting refining data in the ladle bottom blowing refining process at least comprises:
obtaining a slag hole image;
performing binarization processing on the slag hole image to obtain a binarized image;
setting a region with the gray value larger than a preset threshold value in the binarized image as a slag hole, and setting a region with the gray value smaller than or equal to the preset threshold value in the binarized image as a slag layer;
by the formulaCalculating to obtain a slag eye area occupation ratio S; wherein a is the total number of pixels of the slag hole, and b is the total number of pixels of the slag layer.
Specifically, the preprocessing the refined data includes:
deleting abnormal values, missing values and repeated values in the refined data, and carrying out normalization processing on the deleted refined data.
Specifically, after deleting the outlier, the missing value, and the repeated value in the refined data, the method further includes:
performing data expansion on the deleted refined data;
the normalization processing for the deleted refined data specifically comprises the following steps:
and carrying out normalization processing on the refined data after the data expansion.
Specifically, before the preprocessing data is input to a machine learning regression decision tree model for prediction, the method further comprises:
calculating entropy values of all input refined data, taking the input refined data with the largest entropy value as a root node of a decision tree, taking the input refined data with the next largest entropy value as an intermediate node under the root node, and the like until all the input refined data are all nodes of a spanning tree, wherein the decision tree model is established.
The invention also provides a refined ladle slag prediction system adopting the regression decision tree, which comprises:
the refining data acquisition module is used for acquiring refining data in the ladle bottom blowing refining process;
the refining data preprocessing module is used for preprocessing the refining data;
and the prediction module is used for inputting the preprocessed data into a machine learning regression decision tree model for training and predicting to obtain a prediction result.
Specifically, the refining data acquisition module at least comprises:
a slag hole image obtaining unit for obtaining a slag hole image;
the image binarization unit is used for carrying out binarization processing on the slag hole image to obtain a binarized image;
the gray value comparison unit is used for setting a region with a gray value larger than a preset threshold value in the binary image as a slag hole, and setting a region with a gray value smaller than or equal to the preset threshold value in the binary image as a slag layer;
slag hole surfaceA volume ratio calculating unit for passing through the formulaCalculating to obtain a slag eye area occupation ratio S; wherein a is the total number of pixels of the slag hole, and b is the total number of pixels of the slag layer.
Specifically, the refined data preprocessing module includes:
a refined data cleaning unit, configured to delete an outlier, a missing value, and a duplicate value in the refined data;
and the data normalization unit is used for normalizing the deleted refined data.
Specifically, the refined data preprocessing module further comprises:
the data expansion unit is used for carrying out data expansion on the deleted refined data;
the data normalization unit is specifically used for performing normalization processing on the refined data after the data expansion.
Specifically, the method further comprises the steps of:
the decision tree model building module is used for calculating entropy values of all input refined data, taking the input refined data with the largest entropy value as a root node of a decision tree, taking the input refined data with the next largest entropy value as an intermediate node under the root node, and the like until all the input refined data are all generated into nodes of the tree, and representing that the decision tree model building is completed.
One or more technical schemes provided by the invention have at least the following technical effects or advantages:
refining data in the ladle bottom blowing refining process is collected, then the refining data is preprocessed, and the preprocessed data is input into a machine learning regression decision tree model for training and prediction, so that a prediction result is obtained. According to the invention, the refined ladle structure is converted into a corresponding machine learning special structure, unnecessary interaction and coupling are avoided, network parameters with similar characteristics are reused, the information of ladle refining physical principles is introduced, the dependence of a model on a sample is greatly reduced, and the modeling, training and learning efficiency and the model application success rate of a decision tree model can be remarkably improved. According to the invention, the decision tree training model is built for training the water model experimental data and the industrial field experimental data, so that the content of components in the ladle refining process and the condition in the ladle furnace under different conditions are judged, the ladle in the working process can be predicted and monitored in real time, the optimal experimental condition is selected to maximize the secondary purification degree of molten steel, the unexpected condition is predicted and adjusted in time, and the problems of green production, efficiency improvement, raw material energy waste and the like can be well carried out.
Drawings
FIG. 1 is a flow chart of a prediction method for refined ladle slag by adopting a regression decision tree according to an embodiment of the invention;
FIG. 2 is a block diagram of a prediction system for refined ladle slag using a regression decision tree according to an embodiment of the present invention;
FIG. 3 is a schematic diagram of prediction of a method and system for predicting ladle slag of refined steel by adopting regression decision tree according to an embodiment of the present invention;
FIG. 4 is a software interface diagram of a refined ladle slag prediction system constructed in accordance with an embodiment of the present invention.
Detailed Description
The embodiment of the invention can predict and monitor the ladle in real time in the working process by providing the prediction method and the prediction system for the ladle slag of the refined steel by adopting the regression decision tree, thereby effectively guiding industrial production.
The technical scheme in the embodiment of the invention aims to achieve the technical effects, and the overall thought is as follows:
the embodiment of the invention relates to a prediction method and a prediction system for refining ladle slag-rolling behaviors in a steel production process by using a Sklearn regression decision tree algorithm, wherein the prediction method comprises the following steps: establishing a refining ladle refining database, data preprocessing, model parameter setting, data training and data prediction. The method comprises the steps of establishing a refining ladle refining database, mainly obtaining slag hole images through experiments on industrial sites or water model experiments, obtaining mass production data through a plurality of groups of experiments, and establishing the database; the data preprocessing is to screen the data in the database, exclude the data with larger error, process the missing value, and then divide the data set into 80% training data and 20% forecast data; setting model parameters, namely setting data characteristics of input and output layers, adjusting the maximum depth of a tree, random seeds, the number of maximum leaf nodes and the like; the data training is to perform training fitting on 80% of training data input into a program, then test 20% of predicted data, compare a predicted value with an experimental value, and obtain accuracy, error rate, mean square error and average absolute error, so as to judge the prediction capability of the model, and adjust model parameters according to the capability to optimize the model. The data prediction means that after model optimization is completed, experimental result prediction is carried out on input parameters, and industrial production results can be predicted in advance by obtaining prediction results through real-time prediction, so that guidance is provided for industrial production; and finally, packaging and packaging the program by using Python language to form software, and pushing the software to industry for application.
In the embodiment of the invention, all model inputs are divided into two types according to a physical principle, and experimental conditions (bottom blowing position, slag layer thickness, blowing time, blowing flow, power on time, power on quantity and the like) of ladle refining and parameters (data characteristics of input and output layers, maximum depth of a regulating tree, random seeds, maximum leaf node number and the like) of a machine learning regression decision tree model are divided into two types; the outputs are divided into two categories, the experimental results of ladle refining (slag hole area, slag particle size and morphology, temperature, refining ladle lining wear, total oxygen content, etc.) and the evaluation parameters of the machine learning regression decision tree model (mean absolute error (MAE), mean Square Error (MSE), training score (trainescore), test score (TextScore), and decision coefficient (R2)).
In order to better understand the above technical solutions, the following detailed description will refer to the accompanying drawings and specific embodiments.
Referring to fig. 1, the method for predicting refined ladle slag by adopting regression decision tree provided by the embodiment of the invention comprises the following steps:
step S110: collecting refining data in the ladle bottom blowing refining process;
specifically, experimental data generated by a water model experiment or an industrial field refining ladle refining is obtained, wherein the experimental data comprise data such as a slag hole area ratio, the depth of slag particles in molten steel, the size and form of slag particles, the distribution condition of temperature and particle size in a flow field, slag rolling components, abrasion of a refining ladle lining, total oxygen amount and the like, and a refining ladle bottom blowing refining database is established.
The distribution of oil drop particles can be photographed by a high-speed camera at a high speed, and the distribution condition of a flow field and the flow condition of the particles in the flow field can be determined by hundreds or thousands of pictures. When the occupation ratio of the slag hole area is obtained, collecting refining data in the ladle bottom blowing refining process, comprising the following steps:
shooting by using a high-definition camera to obtain a slag hole image;
performing binarization processing on the slag hole image to obtain a binarized image;
setting a region with gray values larger than a preset threshold value in the binarized image as a slag hole, and setting the region as white; setting a region with gray value smaller than or equal to a preset threshold value in the binarized image as a slag layer, and setting the slag layer as black; and setting the areas outside the slag eyes and the slag layer of the image to be black or white respectively to obtain two images after binarization processing. And respectively carrying out scanning statistics on the two pixels according to the pixel number statistics mode, wherein the number of white area (slag eye) pixels in the image with the black background is a, and the number of black area (slag layer) pixels in the image with the white background is b.
By the formulaCalculating to obtain a slag eye area occupation ratio S; wherein a is the total number of pixels of the slag hole, and b is the total number of pixels of the slag layer.
Step S120: preprocessing refined data;
the step is specifically described, and the pretreatment of the refined data comprises the following steps:
deleting abnormal values, missing values and repeated values in the refined data, and carrying out normalization processing on the deleted refined data.
To avoid unbalance of the data to improve the accuracy of prediction, after deleting the outliers, missing values, and duplicates in the refined data, the method further comprises:
performing data expansion on the deleted refined data;
in this case, normalization processing is performed on the deleted refined data, specifically including:
and carrying out normalization processing on the refined data after the data expansion.
Step S130: and inputting the preprocessed data into a machine learning regression decision tree model for training and predicting to obtain a prediction result.
Specifically explaining the step, inputting the preprocessed data into a machine learning regression decision tree model for training and prediction, wherein the method comprises the following steps:
and inputting the preprocessed data into a machine learning regression decision tree model for training until the preset training iteration times are reached, and completing model training.
And inputting the input refined data into the trained machine learning regression decision tree model for prediction to obtain a prediction result.
Describing the establishment process of the decision tree model, before the preprocessed data is input into the machine learning regression decision tree model for prediction, the method further comprises the following steps:
calculating entropy values of all input refined data, taking the input refined data with the largest entropy value as a root node of a decision tree, taking the input refined data with the next largest entropy value as an intermediate node under the root node, and so on until all the input refined data are all nodes of a spanning tree, and indicating that the establishment of a decision tree model is completed.
That is, according to the calculation of the entropy value, the refined data feature with the largest influence factor on the prediction result is taken as the root node of the decision tree, the input refined data is divided according to the difference of the attribute values, the entropy value of other refined data features is compared except the refined data feature of the root node, the input refined data feature with the largest entropy value is taken as the middle node under the root node, and the steps are sequentially circulated until all the nodes of the decision tree are generated by the input refined data feature, so that the establishment of the decision tree model is completed. And (5) observing evaluation indexes such as MAE, R2, trainScore, textScore and the like of experimental results, and judging whether the model parameters are good or not according to the indexes. If the expected value is not reached, the structure of the decision tree needs to be continuously adjusted, namely, parameters such as the maximum depth of the tree, the random seed, the maximum leaf node number and the like are adjusted until the attribute of the decision tree model reaches the optimal value. After the model parameters are adjusted to be optimal and the data training is completed, experimental parameters are input to obtain a predicted result, and the obtained predicted value has instantaneity and can effectively guide industrial production.
In the embodiment of the invention, in the recursion process of the decision tree from root to leaf, a 'partition' attribute is searched for in each intermediate node, and three stopping conditions exist:
1) The sample total attribute contained in the current node belongs to the same category, and no division is needed;
2) The current attribute set is empty, and no other attributes can be continuously classified and cannot be divided;
3) The sample set contained by the current node is empty and cannot be divided.
In the case of predicting slag eye area, experimental conditions were determined as three input refining data: the experimental result is the size of the slag hole area, and the influence degree ratio of the bottom blowing flow, the slag layer thickness and the single bottom blowing hole position to the prediction result is found through the entropy value attribute: 66.047%, 32.311% and 1.641% indicate that the bottom blowing flow plays a decisive role in the size of the slag hole area, and secondly that the thickness of the slag layer and the position of the single bottom blowing hole have almost no influence on the experimental result.
And (3) obtaining a data result display through training:
in the machine learning decision tree model training of predicting the slag eye area, the model prediction accuracy is up to 98.229%, the test result is up to 97.243%, and MAE and R2 values are up to 91.481% and 97.243%, respectively, which shows that the decision tree model provided by the embodiment of the invention has stronger prediction capability.
Referring to fig. 2, the refined steel ladle slag prediction system using regression decision tree provided by the embodiment of the invention includes:
the refining data acquisition module 100 is used for acquiring refining data in the ladle bottom blowing refining process; specifically, experimental data generated by a water model experiment or an industrial field refining ladle refining is obtained, wherein the experimental data comprise data such as a slag hole area ratio, the depth of slag particles in molten steel, the size and form of slag particles, the distribution condition of temperature and particle size in a flow field, slag rolling components, abrasion of a refining ladle lining, total oxygen amount and the like, and a refining ladle bottom blowing refining database is established.
Specifically, the refining data acquisition module 100 includes at least:
the oil drop particle distribution obtaining unit is used for carrying out high-speed shooting through a high-speed camera, and the distribution condition of a flow field and the flow condition of particles in the flow field can be determined by hundreds or thousands of pictures;
a slag hole image obtaining unit for obtaining a slag hole image by shooting with a high-definition camera;
the image binarization unit is used for carrying out binarization processing on the slag hole image to obtain a binarized image;
the gray value comparison unit is used for setting a region with gray value larger than a preset threshold value in the binarized image as a slag hole and setting the slag hole as white; setting a region with gray value smaller than or equal to a preset threshold value in the binarized image as a slag layer, and setting the slag layer as black; and setting the areas outside the slag eyes and the slag layer of the image to be black or white respectively to obtain two images after binarization processing. And respectively carrying out scanning statistics on the two pixels according to the pixel number statistics mode, wherein the number of white area (slag eye) pixels in the image with the black background is a, and the number of black area (slag layer) pixels in the image with the white background is b.
A slag eye area ratio calculating unit for calculating the slag eye area ratio according to the formulaCalculating to obtain a slag eye area occupation ratio S; wherein a is the total number of pixels of the slag hole, and b is the total number of pixels of the slag layer.
A refined data preprocessing module 200, configured to preprocess refined data;
specifically, the refined data preprocessing module 200 includes:
the refined data cleaning unit is used for deleting abnormal values, missing values and repeated values in the refined data;
and the data normalization unit is used for normalizing the deleted refined data.
To avoid unbalance of the data to improve accuracy of the prediction, the refined data preprocessing module 200 further includes:
the data expansion unit is used for carrying out data expansion on the deleted refined data;
in this case, the data normalization unit is specifically configured to normalize the refined data after the data expansion.
The prediction module 300 is configured to input the preprocessed data to a machine learning regression decision tree model for training and predicting, so as to obtain a prediction result.
Specifically, the prediction module 300 includes:
the model training unit is used for inputting the preprocessed data into the machine learning regression decision tree model for training until the preset training iteration times are reached, and model training is completed.
And the prediction unit is used for inputting the input refined data into the trained machine learning regression decision tree model for prediction to obtain a prediction result.
Describing the establishment process of the decision tree model, the method further comprises the following steps:
the decision tree model building module is used for calculating the entropy value of each input refined data, taking the input refined data with the largest entropy value as the root node of the decision tree, taking the input refined data with the next largest entropy value as the intermediate node under the root node, and so on until all the input refined data are all generated into the nodes of the tree, and representing that the decision tree model building is completed. That is, according to the calculation of the entropy value, the refined data feature with the largest influence factor on the prediction result is taken as the root node of the decision tree, the input refined data is divided according to the difference of the attribute values, the entropy value of other refined data features is compared except the refined data feature of the root node, the input refined data feature with the largest entropy value is taken as the middle node under the root node, and the steps are sequentially circulated until all the nodes of the decision tree are generated by the input refined data feature, so that the establishment of the decision tree model is completed. And (5) observing evaluation indexes such as MAE, R2, trainScore, textScore and the like of experimental results, and judging whether the model parameters are good or not according to the indexes. If the expected value is not reached, the structure of the decision tree needs to be continuously adjusted, namely, parameters such as the maximum depth of the tree, the random seed, the maximum leaf node number and the like are adjusted until the attribute of the decision tree model reaches the optimal value. After the model parameters are adjusted to be optimal and the data training is completed, experimental parameters are input to obtain a predicted result, and the obtained predicted value has instantaneity and can effectively guide industrial production.
In the embodiment of the invention, in the recursion process of the decision tree from root to leaf, a 'partition' attribute is searched for in each intermediate node, and three stopping conditions exist:
1) The sample total attribute contained in the current node belongs to the same category, and no division is needed;
2) The current attribute set is empty, and no other attributes can be continuously classified and cannot be divided;
3) The sample set contained by the current node is empty and cannot be divided.
The following describes embodiments of the present invention in more detail:
referring to fig. 3, an embodiment of the present invention mainly includes: establishing a database, preprocessing data, setting model parameters, training data and predicting data. The database mainly comprises data acquired in the refining ladle industrial field production and data acquired in the water model experiment, the image data is processed through binarization, and the area calculation is performed by using Python language development software, so that the slag hole area data can be obtained. The data preprocessing is to compare and filter all the data in the database, delete the data with repeated data and larger error, convert the data into a suitable text form, and be convenient to store in excel or csv files. Carrying out Min-Max normalization on the data to carry out linear transformation, mapping the data to intervals [0,1], limiting the preprocessed data in a certain range, eliminating adverse effects caused by singular sample data, and dividing a data set into: 80% training data and 20% test data, the dataset was converted to dictionary output form by dictionary specific engineering. The model is optimized through training data, the adaptability of the model to the data can be verified through test data, and the error is overlarge due to poor model training effect. The model parameter setting is to set parameters on the basis of a built decision tree algorithm model, wherein the model parameter setting comprises the neuron numbers of each layer of an input layer, a hidden layer and an output layer, the number of layers of the hidden layer, and the setting input characteristics comprise: bottom blowing position, slag layer thickness, blowing time, blowing flow, power-on time, power-on quantity and the like; the output characteristics include: slag hole area, slag coiling depth, grain size distribution, slag coiling components, abrasion of lining of refining ladle, total oxygen amount and the like. Also including setting an activation function; training iteration number error: the training is stopped when the training iteration times are reached; minimum error: if the predicted value and the experimental value are smaller than the minimum error, training is stopped, and the setting of parameters plays a decisive role in the training function of the model. The data training is a recursion process that after model parameters are adjusted, a decision tree reads data into a program from root nodes to leaf nodes, a partition attribute is searched for at each node, a most characteristic is selected to construct the root node at the beginning of the partition of the decision tree, all training data are placed on the root node, so that a training data set is divided into subsets according to the characteristic, and the subsets enter the child nodes; the process is looped through, and all subsets are recursively partitioned according to the attributes of the internal nodes. If the subsets can be classified substantially correctly, leaf nodes will be constructed and the subsets will be divided into corresponding leaf nodes. 80% of training data in the database is used for model training, 20% of testing data is used for testing models, comparison is carried out according to evaluation indexes, and an optimal machine learning decision tree model is screened for prediction. The data prediction refers to applying an optimized machine learning decision tree model to practice, obtaining corresponding experimental results through different experimental conditions, inputting different condition parameters of an experiment in the model, and obtaining the corresponding experimental results through the model prediction. Specifically, according to the model of the refining ladle and the control conditions, a subset of model inputs are selected by actual inputs; the actual output selects a subset of the model outputs; different input variables are selected as input parameters according to different output characteristics. FIG. 4 is an interface diagram of a prediction software of a machine learning decision tree for refining ladle slag, wherein machine learning codes are written through Python language, and a Thinker database and a Cv2 database are utilized for packaging to form the prediction software of refining ladle slag, and the software comprises four blocks of network parameter setting, data reading, model training and model prediction, so that the functions in a prediction model flow chart of the machine learning decision tree for refining ladle slag are realized, repeated or complicated model training and testing can be automatically executed, the working efficiency of a user is improved, the occurrence of errors is reduced, and the time is saved.
The embodiment of the invention provides a prediction method and a prediction system for applying a machine learning regression decision tree algorithm to refined ladle slag, which can predict and monitor a ladle in real time in the working process, effectively solve the technical problem that the traditional metallurgy cannot make real-time judgment on industrial field production, and can well perform green production, improve efficiency, avoid raw material energy waste and the like through the real-time prediction of the machine learning decision tree model.
It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as a method, system, or computer program product. Accordingly, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
Embodiments of the present invention are not described in detail and are well known to those skilled in the art. Finally, it is noted that the above embodiments are only for illustrating the technical solution of the present invention and not for limiting the same, and although the present invention has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that modifications and equivalents may be made thereto without departing from the spirit and scope of the technical solution of the present invention, which is intended to be covered by the scope of the claims of the present invention.
Claims (10)
1. A prediction method for refined ladle slag by adopting a regression decision tree is characterized by comprising the following steps:
collecting refining data in the ladle bottom blowing refining process;
preprocessing the refined data;
and inputting the preprocessed data into a machine learning regression decision tree model for training and predicting to obtain a prediction result.
2. The method for predicting ladle slag inclusion in a refined steel using a regression decision tree of claim 1, wherein the collecting refining data from the ladle bottom blowing refining process comprises at least:
obtaining a slag hole image;
performing binarization processing on the slag hole image to obtain a binarized image;
setting a region with the gray value larger than a preset threshold value in the binarized image as a slag hole, and setting a region with the gray value smaller than or equal to the preset threshold value in the binarized image as a slag layer;
by the formulaCalculating to obtain a slag eye area occupation ratio S; wherein a is the total number of pixels of the slag hole, and b is the total number of pixels of the slag layer.
3. The method for predicting ladle slag in a refined steel using a regression decision tree of claim 1, wherein said preprocessing the refined data comprises:
deleting abnormal values, missing values and repeated values in the refined data, and carrying out normalization processing on the deleted refined data.
4. The method for predicting ladle slag in a refined steel using a regression decision tree according to claim 3, further comprising, after said deleting the outlier, the missing value, the repeated value in the refined data:
performing data expansion on the deleted refined data;
the normalization processing for the deleted refined data specifically comprises the following steps:
and carrying out normalization processing on the refined data after the data expansion.
5. The refined steel ladle slag prediction method employing a regression decision tree according to any one of claims 1-4, further comprising, prior to said inputting said preprocessed data into a machine learning regression decision tree model for prediction:
calculating entropy values of all input refined data, taking the input refined data with the largest entropy value as a root node of a decision tree, taking the input refined data with the next largest entropy value as an intermediate node under the root node, and the like until all the input refined data are all nodes of a spanning tree, wherein the decision tree model is established.
6. A refined steel ladle slag prediction system employing a regression decision tree, comprising:
the refining data acquisition module is used for acquiring refining data in the ladle bottom blowing refining process;
the refining data preprocessing module is used for preprocessing the refining data;
and the prediction module is used for inputting the preprocessed data into a machine learning regression decision tree model for training and predicting to obtain a prediction result.
7. The refined steel ladle slag prediction system using a regression decision tree of claim 6, wherein the refined data acquisition module comprises at least:
a slag hole image obtaining unit for obtaining a slag hole image;
the image binarization unit is used for carrying out binarization processing on the slag hole image to obtain a binarized image;
the gray value comparison unit is used for setting a region with a gray value larger than a preset threshold value in the binary image as a slag hole, and setting a region with a gray value smaller than or equal to the preset threshold value in the binary image as a slag layer;
a slag eye area ratio calculating unit for calculating the slag eye area ratio according to the formulaCalculating to obtain a slag eye area occupation ratio S; wherein a is the total number of pixels of the slag hole, and b is the total number of pixels of the slag layer.
8. The refined steel ladle slag prediction system using a regression decision tree of claim 6, wherein the refined data preprocessing module comprises:
a refined data cleaning unit, configured to delete an outlier, a missing value, and a duplicate value in the refined data;
and the data normalization unit is used for normalizing the deleted refined data.
9. The refined steel ladle slag prediction system using a regression decision tree of claim 8, wherein the refined data preprocessing module further comprises:
the data expansion unit is used for carrying out data expansion on the deleted refined data;
the data normalization unit is specifically used for performing normalization processing on the refined data after the data expansion.
10. The refined steel ladle slag prediction system employing a regression decision tree of any of claims 6-9, further comprising:
the decision tree model building module is used for calculating entropy values of all input refined data, taking the input refined data with the largest entropy value as a root node of a decision tree, taking the input refined data with the next largest entropy value as an intermediate node under the root node, and the like until all the input refined data are all generated into nodes of the tree, and representing that the decision tree model building is completed.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311438276.3A CN117593254A (en) | 2023-10-30 | 2023-10-30 | Refined ladle slag prediction method and system adopting regression decision tree |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311438276.3A CN117593254A (en) | 2023-10-30 | 2023-10-30 | Refined ladle slag prediction method and system adopting regression decision tree |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117593254A true CN117593254A (en) | 2024-02-23 |
Family
ID=89910587
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311438276.3A Pending CN117593254A (en) | 2023-10-30 | 2023-10-30 | Refined ladle slag prediction method and system adopting regression decision tree |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117593254A (en) |
-
2023
- 2023-10-30 CN CN202311438276.3A patent/CN117593254A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108764517B (en) | Method, equipment and storage medium for predicting change trend of silicon content in molten iron of blast furnace | |
| Chen et al. | Energy consumption modelling using deep learning technique—a case study of EAF | |
| CN112819802B (en) | Method for supervising and predicting blast furnace condition abnormality based on tuyere information deep learning | |
| CN111651931B (en) | Blast furnace fault diagnosis rule derivation method based on deep neural network | |
| CN114678080B (en) | Converter end point phosphorus content prediction model, construction method and phosphorus content prediction method | |
| CN110781944B (en) | Automatic molten iron slag-off control method based on deep learning | |
| CN116934220A (en) | Intelligent warehouse layout method based on intelligent data analysis and algorithm optimization | |
| CN112465223A (en) | Blast furnace temperature state prediction method | |
| CN111915080A (en) | Raw fuel cost optimal proportioning method based on molten iron quality constraint | |
| CN115795380A (en) | Flue gas acid making data cleaning and optimizing method based on isolated forest and weighted random forest | |
| Nenchev et al. | Metallurgical data science for steel industry: A case study on basic oxygen furnace | |
| CN117593254A (en) | Refined ladle slag prediction method and system adopting regression decision tree | |
| Sala et al. | Hybrid static-sensory data modeling for prediction tasks in basic oxygen furnace process | |
| CN113177353A (en) | Data model construction method applied to industrial early warning system | |
| CN111444963A (en) | Blast furnace molten iron silicon content prediction method based on SSA-SVR model | |
| WO2022059753A1 (en) | Method for generating trained prediction model that predicts energy efficiency of melting furnace, method for predicting energy efficiency of melting furnace, and computer program | |
| CN114896228B (en) | Industrial data stream cleaning model and method based on filtering rule multistage combination optimization | |
| CN115094193A (en) | Intelligent molten iron pretreatment desulfurization system based on data mining | |
| Cui et al. | Prediction of blast furnace temperature based on multi-information fusion of image and data | |
| CN112861276A (en) | Blast furnace burden surface optimization method based on data and knowledge dual drive | |
| CN112183642A (en) | Method and system for detecting coal consumption of cement firing based on random forest model | |
| CN117766065A (en) | Converter optimization pushing method based on industrial big data | |
| Rao et al. | Automation of metal charge calculations using support vector machine | |
| Ge | A neural network approach to the modeling of blast furnace | |
| Yang et al. | The modeling of time series based fuzzy cognitive maps and its application in the furnace temperature prediction |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |